CA1145928A - Method and apparatus for grading fruit - Google Patents

Abstract

ABSTRACT

This invention relates to apparatus for auto-matically grading and sorting articles, especiallg fruit, according to size surface blemish and surface color.
Prior art automatic processes have resulted in incomplete grading and are incapable of providing certain important information such as abrupt variations of blemishes and size related information. The present invention provides an improved apparatus and method for grading and sorting articles, particularly fruit, according to size, surface blemish and surface color. Fruit is passed sequentially through a camera array which scans the surface of each fruit and measures the intensity of light reflected from successive discrete surface segments. Significant differ-ences between such measured intensities are detected and a measurement of surface blemish is generated in accor-dance therewith. Size measurements are derived by count-ing the total number of segments in the surface of each fruit. Color measurements are derived by averaging the ratio of red light intensity to infrared light intensity reflected from each of a plurality of surface areas of each fruit. The fruit are separated and delivered to separate receivers by a mechanism responsive to the size; blemish and color measurements of the respective fruit.

Description

5~Z8 ~
MEIHOD AND APPARATUS FOR ~DING FRUI'r TECHNICAL FIELI) The present invenki~n relates ~enerally to ~ort-ing apparatus and~ more particularly, to apparatus ~or automatically grading and sorting articles~ especially fruit, according to size, sur~ace blemish and sur~ace color.
BACK~ROUNO OF THE PRIOR ART
., The grading and ~orting of fruit is a ma~ or cost factor ~or the ~re~h ~ruik industry. In the past, mo~t grading and sorking has been per~ormed by human labor, involving the visual inspectlon o~ each ~ruit and the manual depositing of such ~ruit into a number of separ~
ate recelvers in accordance with a worker's asses~ment of the ~ruit's proper grade category.
In addition to being a slow process, manual grad-ing and sorting o~ fruit has proven to be ~urther de~icient in that the workers' grading assessments are hlghly sub~ectiveg varying both with time and from worker to worker. Moreover, a single blemish or dis-colored area on one side of a ~ruit can occasionally escape detectlon during manual sorting.
Beca se of these de~iciencies in the manual grad-ing and sorting of fruit, there have been a number of aktempts in the past to automate the grading and sortlng process. Studies have been made, su¢h as that d~scribed in U. S. Patent No. 2,933,613 to J. B. Power~ entitled "Method and Apparatus ~or Sorting Ob~ects According to Color", which indicate that a measure o~ khe sur~ace color o~ ~ruit can be derived by computing a ratio o~
the intensity of reflected llght having a ~ir~t wave-length to the intensity of reflected light havlng a second wavelength. Accordingly, devices have been con-strucked and used for measuring the ratlo of red light 3.5 in~enslty to in~rared light intensity received~from the fruit sur*ace. However, such devices have typically provlded only a single measurement for each fruit, and have done so by inspecting only one side of the ~ruit.
Since ~ruit can typlcaIly have contrasting colors for different portions of their sur~aces, these devices have :- . `', not been entirel-y success~ul.
Other studies have been made, such aæ that de-~crlbed ln United States Patent No. 3g867,041J to G. K.
Brown et al entitled "Method for Detectlng Bruises ln Fruit';, which indlcate that bruised fruit reflect light to a markedly less degree than do unbruised ~ruit.
Typical fruit grading dev~ces that utilize this princl-ple, however, make only a single measurement of the intensity of llght reflected from the surface of the frult. The devices do not detect abrupt variation~ in the reflect~vity of the fruit sur~aceJ such as those commonly exhibited by sur~ace blemishes in fru~t~
especially citrus fruit. Addltionallyg successful per-formance of such prlor devices requires maintenance o~ a 15 constant level of illumination, a requirement that is difficult to achieve ln the environment in which such devices are typically used.
me sorting Or fruit according to size has usually been performed in the past either by manual inspection or by a separate automatic slzing apparatus. This has necessitated multiple inspections of each fruitJ thus aggravating the inefficiencies and performance draw~
back~ of such prior fruit sortlng systems.
It will be appreciated from the foregolng that 25 there is a definite need ror a more reliable and more efflcient technique ~or grading and sortlng ~ruit accord-ing to sizeJ blem-lsh and color. In particular such a technique should utilize apparatus that perform~ merely one inspection of substantlally the ent~re ~urface of 30 each fruit, and should have sufficient resolution to detect even minute blemishes or flaws in the fruit surface and to allow grading into a relatlvely large number of categories. The present lnvention ful~llls thi~ need.
BRIEF SUMMARY OF THE INVENTION
. .
The present lnventlon is embodled in a method and apparatus for gradlng and sorting artlcles; eRpecl- :
ally frult, according to size, surface color and sur~ace blemish. In accordance ~ith the inventlon, the apparatus includes camera means f'or sensing light re~lected from - the sur~ace o~ each fruit and generating a plurallty of corresponding light measurement signal~, which are transmitted to blemish detection c-lrcuitry ~or detecting signlficant variations between them, to ob~ain a measure of the degree of blemlsh on the surface of each fruit.
Additionally the light measurement signals are substan-tially concurrently transmitted to color detection cir-cuitry for obtaining a color measurement for each of several di.stinct areas on the sur~ace of each ~ruit.
More particularlyg the subject apparatus include~
a conveyor for continuously movlng ~ruit one by one through an examining region where each fruit ls ex-amined sequentially by the camera means. The camera means includes a number of scanning or segmental cameras for generating light measurement signals that are trans-mitted to and processed by the blemish detection cir-cuitry, and in addltiong includes a number of separate color~sensltive cameras for generating other light measurement signals that are transmitted to and pro-cessed by the color detection circuitry.
The segmental cameras are circumferentially arranged in a blemish examinlng plane through which the ~ruit to be examined and graded is passed. Similarly~
the color-sensitive cameras are circumferentlally arranged in a color examining plane through which the fruit is passed. The frllit is uni~ormly illuminated as it is dropped through the blemlsh examining and color examining planes, to provide light input to the ~eg-3 mental and color-sensitive cameras.
In the preferred embodiment of the inventiong each segmental camera includes a llnear array of photo-diodes, located in the blemish examinlng plane and sub-~tantially circumferential with respect to a central region of the plane through which the fruit is passed.
When a frult is passing through the plane, each photo-diode will receive reflected light from a uni~ue seg-ment o~ the fruit surface, and will generate an elec-trical signal proportional to the intensity o~ the light ' --4 ~
received from that segment The elect~ical signal~ from all o~ th~ photo-diodes are read in a cyclic sequence, wlth the signals from the photodiodes of each segmental camera belng read only a~er those generated by the photodlodes o~ the previous segmental ca~era. Since the fruit will have moved an incremental distance through the blemish ex-amining plane during the time taken to read the signals ~rom all of the photodiodes in one full cycle, it will be apparent that repetition of the sequential readlng cycle will provide scans of additional, approxi~ately pl~nar portions of the fruit surface. In this ~anner, substantially the entire fruit surface can be examined by the photodiodesg in a helical scanning fashlon.
The cyclic sequence of electrical signals de-rived from the photodiodes is designated a sequentlal scan signal, and~ in accordance with one aspect of the invention, each successive value in this signal is compared, for example by divisiong with the values for neighboring segmentsg and a sequential correlation sig-nal is generated in accordance with the comparisons made. This sequential correlation signal represents a measure of irregularities in the reflectivity of the fruit surfaceg such irregularities being due prlmarily to surrace blemlshes.
The correlation signal is then filtered to BUb-~tantially ellminate all slowly varying signal ¢om-ponents not attributable to surface blemishes, such as those caused by the curvature of the ~ruit. The fil-tered correlation signal is khen further processed inan absolute value detector so that both positlve and negative variations in surface reflectiYity are taken into account. Finally9 an in~egrator to which the re-aultant slgnal is fed~provides a measure of the total surface~blemish of the ~ruit.
A measure of the size o~ each ~ruit is obtained b~ counting the number of segments detected in the sur-~ace of the ~ruit as lt passes the blemish examining plane. By dividing the measure of total surface blemish `~,'.

on the fruit by this measure of si~e, a normalized measure of the degree of sur~ace blemish can be obtained.
In order to detect fruit color, each color-sensitive camera in the apparatus of the invention in-cludes a red phototransducer and an infrared photo-transducer. Reflected light received by each of the cameras in the color e~amining plane is first directed at a beam splitter. One portion of light from the beam splitter is passed through a red light filter before reaching the red phototransducer, and an equal portion is passed throu~h an infrared light filter before reaching the infrared phototransducer. In this manner, each phototransducer in the pair receives light from the same portion of the fruit as it passes through the color examining plane.
More specifically, each color phototransducer generates an output signal indicative of the intensity of light incident on it. In accordance with one aspect of the invention, the output signals from each photo-transducer pair are read in a se~uential fashion andthe measure of red light intensity is compared, for example by division, to the measure of infrared light intensity for each pair. Since the magnitude of re-flected infrared ligh-t does not vary substantially with fruit ripeness or color, while the magnitude of reflected red light does so vary, the comparison (e.g.
ratio) of the two signals is an effective measure oE
the color of a fruit.
Duriny the time taken to measure the output signals from each phototransducer pair, and to compute the ratios of such signals, the fruit being e~amined will have moved an incremental distance through the color examining plane. The phototransducers, then, will provide output signals corresponding to the reflected light intensities for different portions of the fruit.
Repeating -the sequential phototransducer reading and ratio computation as the frui moves completely through the examining plane provides color information for sub-stantially the entire fruit sur~ace.

, ::

3'~

The separately obtained color ratios ~or each fruit are then numerically averaged, to aerive ~ measure of the average color of the fruit surface. Addition-ally, the separate color ratios are compared to a pre-determined threshold and color count pulses are producedwhenever the threshold is exceeded, or alternately, not exceeded. By counting the number of color count pulses for each fruit, measures of the amount of surface having a prescribed color are produced.
The measurements of normalized surface blemish, surface size and surface color, all obtained from the apparatus of the invention/ are utilized to assign each fruit to a particular category or grade. The means employed to so assign the fruit can take any of a wide variety of specific forms, but can most conveniently take the form of a hard-wired or programmable computer.
Control signals provided by such a computer are utilized to actuate appropriate solenoids, and thereby discharge the fruit to particular receivers in accor-dance with the grade determinations. An example ofapparatus for accomplishing this sorting process can be found in United States Patent Nos. 3,768,645 to T.D.
Conway, et al, entitled "Method and Means for Automatic-ally Dectecting and Sortina Produce According to Internal Damage", and 3,930,994, also issued to T.D. Conway et al, and entitled "Method and Means for Internal Inspec-tion and Sorting of Produce".
It will be apparent from the foregoing summary that the present invention represents a significant advance in apparatus and methods for grading fruit. In particular, the apparatus o the present invention grades fruit according to surface blemish, surface color, and size, and does so simultaneously by scanning substantially the entire surface of the fruit. Many other advantages and features of the present invention will become apparent from the following more detailed description of a preferred embodiment, taken in conjunc-tion with the accompanying drawings, which disclose, by way of example, the principles of the invention.

"t~2~

BRIEF DESCRIPTION OF ~IE DRAWIN~S
FIG. 1 ls a slde elevational view o~ a ~rult transport structure in which the apparatus o~ the present lnvention ls employed; showing in particular the ~ruik conveyors, the camera array and the sorting station;
FIG. 2 is a plan view of the camera array, taken substantially along the line 2~2 in FIG. l;
FIG. 3a is a simplifled sectional view o~ a segmenkal camera and a color-sensitive camera, taken ~ ~ubstantiall~ along the line 3a-3a in FIG. 2;
FIG. 3b is a simplified perspective and schematiG
view ~ a se~mental and color-sensitive camera pair, ~howing the paths of light reflected from a fruit in the examining region to the respective cameras;
FIG. 3c is a simplified block di.agram of the circuitry of a fruit grading apparatus constructed in accordance with the present invention;
FIG. 4 is a more detailed block diagram of the fruit grading apparatus of FIG. 3c~
FIG. 5 is a more detailed block diagram of the camera and signal formatter circultry of the apparatus of FIG. 4;
FIG. 6 is a more detailed block diagram of the demultiplexer o~ the apparatus of FIG. 4;
FIG. 7 is a more detailed block diagram ~f the blemish detection circuitry of the apparatus of FIG. 4;
FIG. 8 is a diagrammatical representation o~ the composite YieWS seen by the four segmental cameras as a ~ruit drop~ through their fiolds of view;
3 FIG. 9 is a more detailed view o~ a portion of the composite view ~ one segmental camera in FIG. 8, FIG. 10 is a diagrammatical Yiew of a portion o~
the camera scan signal for one segmental camerag super-imposed on a blemished portlon of a fruit surface to which it corresponds;
FIG. 11 is a simplified schematic diagram of the scan select circuit of the blemish detection circuitry o~ FIG. 7;
FIG. 12 is a more detailed block diagram of th~

': ' ' . : , ,- ':

high pass filter o~ the blemish deteckion circuitry o~
FIG. 7;
~IG. 13 is a simplified flow diagram of one filter section of the hi~h pa~s,~ilter of FI~. ll;
FIG. 14 (sheet 7 with Fig. 6) is a simplified schematic diagram of the blemish on/off timing circuit of the blemish detection circuitry of FIG . 7 ~
FI~. 15 is a more detailed block diagram of the color detection circultry of the apparatus of FIG. 4, lO and FIG. 16 is a flowchart showingg in simpllfled for~g the operational steps performed by a computer in processing blemish9 color and size measurements derived by apparatus of the present invention.
DETAILED DESCRIPTION OF THE_ INVENTION
l. Ove~view .
As shown ~n the exemplary drawingsg the present invention is embodied in an improved apparatus for gradlng and sorting fruit according to size9 surface 20 color and surface blem~sh. It will be understood that, while the invention is partlcularly well suited for detecting surface blemishes~ surface color and size of fresh fruit~ it could be used just as effeckively for the detect~on of irregularities in surface re~lectlvity, 25 color and size of other like artlcles.
In accordance with the present inventlong ~ruit 21 are recelved on a first conveyor 23 and are passed one by one through a camera array 25 that includes seg-mental cameras 31 used in detecting the degree~ if any, to which the surface of each ~ruit is blemished, and color-sensitive cameras 33 used ln determining the aver-age color of each frult, The segmental cameras 31 measure the intenæity of light reflected ~rom each of a plurality o~ segments of the surface of each fruit 21, and the color-sensitive cameras 33 measure tlie intensitles of both red and infrared light reflected from a plurality of narrow strips on the fruit surface. As shown generally in FI~. 3c, signals from the ~egmental cameras 31 and the .. .

, Z~
_9 _ color-sensitlve cameras 33 are suitably multl~lexed together in camera and signal formatter circuitry 32~
wh-lch, in turn9 transmits the multiplexed signals to a demultiplexer 34. The demultiplexer 34, which may be conveniently located in a remote control room, separates the data slgnals and transmits them to blemlsh detec-tion circuitry 35 and color detection circuitry 39~ for further processing.
The blemish detection circuitry 35 receives 10 measurement signals generaked in the segmental cameras 31 and compares the signal for each segment to corres-ponding signals for neighboring segments, to obtain a comparlson or quotient se~uence signal ~ndicative of the degree of irregularlty in reflectivity of the surface of the fruit. The blemish detection circuitry 35 also filters and integrates the quotient sequence signa, to ob~ain a measure of the total surface blemish for each fruit. Simultaneously~ a size detection circu~t 37 (FIG. 7), lntegral with the blemish deteckion circuitry 35, counts the number of segments in the total reflec-tive sur~ace of each fruit, to obtain a measure of the size of the ~ruit.
The color detection c~rcuikry 39 receives signals derived from the color-sensitive cameras 33~ and sums togekher ratios of red llght intenslty to infrared l-lght intensity for each of the narrow strips on the surface o~ each fruit. A count of the number of narrow strips on the surface of each fruit is simultaneously genera-ted.
3 The successive measures of surface blemish and fruit size from the blemlsh detection circuitry 35, and the successlve color ratio summations and counts of surface strips from the color deteckion circuitry 39, are transmltted to a computer 40, which generates normalizsd measures of surface blemish by dividin~ the successive measures o~ total surface blemish by the corresponding measures of ~ruit slze. Slmultaneously the computer 40 determ~nes the average color o~ each ~ruit by dividlng the successive color ratio summations by the corre~ponding count~ of ~urface strlp~, In accordance with the successive normalized measures o~ sur~ace blemish and average color de~e~mina-tions, the computer 40 provides control signals to app~o-priate solenoids 27 at a sorting station 28, to dlvertthe fruit to appropriate locations.

2. Fruit Transport Structure FIG. 1 shows the fruit transport structure that conveys the fruit 21 past the camera array 25 to the sorting station 28, where it is dlverted to speclfled locations by the solenoids 27. Fruit 21 is delivered on the first conveyor 239 with each fruit in a soparate tray 41, to the camera array 25g through which it is dropped in a sequential fashion. Thereafter, the fru~t is received by a second conveyor 29, which compr~ses a series o~ deformable cushions 43s such as bean bags, that move synchronously with the trays 41 of the flrst conveyor 23. The second conveyor used herein is de-scribed in Unlted States Patent NoO 3g961,701 to P. F.
Paddock et al, entitled "Method of and Conveyor ~or Tranaporting Fragile Ob~ects". Each fruit that drops through the camera array 25 ls caught and retained by a single cush~on 43g which then transports the frult to the cortlng statlon 28.

3 ~
As shown in FIG. 2g the camera arra~ 25 houses the ~egmental cameras 31 and the color-sensitlve cameras 33, whlch examine the frult 21 simultaneously as it passes through the array's ~ield of viéw. Add~tionally~
3 the camera array houses illumlnators 45 for directing llght at the fru-lt as it is being examined.
More particularlyg the camera array 25 comprises a donut-shaped carriage in which are housed ~our seg-mental cameras 31~ ~our color-sensitive cameras 33 and ~our broadband illuminators 45. The llluminators are spaced about 90 apart, ln a substantially planar arrangement~ directing light at a centrally located examlning region 47~ through wh~ch the fruit is dropped.
Substantlally the entire sur~ace of the ~ruit is :

illuminated as lt drops through the region. The back-ground area o~ the examinlng region 47 is black and subs~antially non-reflectlveg so that the presence of a fruit can be more readily dekected.
The four segmental cameras 31 are spaced circum-ferentialllJ around the fruit examining region 47. The fields of view of the cameras 31 form a blemish examin-ing plane 49 (FIG. 3a) that is wlthin the fruit examinlng region 47 and substantially perpendicular to the direc-tion of travel of the fruit through it. The segmental cameras 31 are also spaced about 90 apart, with each camera staggered midway between two adjacent illumina-tors 45. The field of view of each segmental camera is sufficlent to permit a full examination of the largest fruit that is to be graded.
Similarly, the four color-sensitive cameras 33 are also spaced circum~erent~ally around the fruit examining region 47J forming a color examining plane 51 (FIG. 3a) that is also within the region 47 and sub-stantially perpendicular to the directlon o~ travel ofthe fruit through it. The color examining plane 51 ~s substantially parallel to the blemish examinlng plane 49, and preferably closely spaced thereto. The color cameras 33 are located at the same angular positions as the segmen~al cameras ~1, in a skaggered relatlonship wlth the illuminators 459 and the field of view o~
each color camera is sufficient to permlt a ~ull exam-ination of the largest frult that is to be graded.
The camera array 25 is equipped with ad~ustable 3 mountlng means 53j as depicted in FIG. 1, whereby the position o~ the array relative to the conveyors 23 and 29 can be adjusted to center the ~alling fruit 21 in the middle of the blemish examining plane 49 and the color examinlng plane 51~ where they can be most effectlvely viewed by all of the segmental and color-sensltive cameras 31 and 33.
As shown in FIGS. 2 and 3b, a heat absorbing fllter 55 is located in front of each llluminator 45.
Thls reduces the intensit~y of llght having wavelengths ~ ' ' , , 3Z~ j beyond the near-infrared band~ thereby reduclng temper~
ature buildup in the fruit examining region 47.
Also located in front o~ the illuminators ~5 i5 a first set of polarizers 57 for allowing transmission of light having only one polarity. Located in front of the segmental cameras 31 and the color cameras 33 is a second set o~ polarizers 59 ~or allowing transmission of only light having the opposite polarity. In this manner, all direct reflections; i.e. 3 ~'glare'l~ ~rom the lO fruit being examined are eliminated from the fields of view of the cameras, and a more authentic indication of the color and reflectivity of the fruit can be ob-talned.
Cooling fans 61 are located ad~acent to each of 15 the il~uminators 45g to dissipate heat generated in the camera array structure9 particularly in the heat ab-sorbing filters 55 and the first set of polarizers 57.
In the illustrated embodiment of the invention9 each of the fans 61 is located between an llluminator 45 and a 20 pair of the cameras 31 and 33, and is oriented to blo~
cooling air across the filter 55 and polarizer 57.
3.l Segmental Camera As showr in FIGS. 3a and 3b~ each segmental camera 31 includes a llnear-photodiode array 63, ~uch 25 as Model No. RLC 6l~P manufacturec1 by Reticon Corporatlon o~ SunnyvaleJ Californla The photodiode array 63 is ~en~itive over a broad range of light wavelengthsg and iB oriented with its axis substantlally perpendlcular to the direction Or travel of the ~ruit~ i.e. 5 with each element of the array positioned to receive light from a different segment of the fruit surface. It will be apparent from FIGS. 3a an~ 3b that each segmental camera 31 is housed with a corresponding color-sensitive camera 33, and that the pair of cameras is protected from dust par~icle contamination by a cover plate 73. Light is received from the examining region 47 through a seg-mental camera aperture 75 located in the cover plate 73, and is focused by a segmental camera lens 77 on the photodiode array 63. Thus the field of view o~ each -- - . : . ,:, ...................... .

.
' - -photodiode array ls a narrow swath o~ the examLning region, substantially perpendicular ~o the direction of travel of the fruit.
3.2 Color-Sensitive_Camera Each color-sensitive camera 33 includes a red phototransducer 65 and an in~rared phototransducer 67.
Each of the red and infrared phototransducers is a conventional di~fused silicon photodiode, such ?~ a PIN-6DP manufactured by United Detector Technologyg Inc.
of Santa Monlcaj California. As shownin FIGS o 3a and 3b, the red phototransducer 65 receives light through a red light filter 695 and thereb~J measures the intensity of red light received by the camera7 and the infrared phototransducer 57 receives light through an infrared light filter 71~ and thereb~T measures the intensity of infrared light received by the camera.
Since the measure of surface color of the fruit being examined is obtained by computing the ratio of red light intensity to in~rared light intel1sity~ it is pre~erable that the red and infrared phototransducers 65 and 67 in each color-sensitive camera 35 have a common fiel~ of view and thus receive light from gener-ally the same source. Preferably, this is accompli6hed using a single color camera lens 78 and a beam splltter 7g, pre~erahly o~ a conventional cube type.
Light is received from the examining region 47 through a color camera aperture 76 located in the cover p~ate 73, and ls focused by the lens 78 through the beam splitter 79 and the respective red and infrared filters 69 and 71, and onto the respectlve red and in~
frared phototransducers 65 and 67. Additionallyg a red phototransducer aperture 81 and an infrared phototrans-ducer aperkure 82j oriented substantially perpendicular to the direction of the ~ruit's travel3 are located in front of the respective red and infrared phototrans-ducers 65 and 670 Each aperture restricts the light incident on such phototransducers to that received from a narrow swath of the examining region 47. Thus, when a fruit ls in the examining region~ the two photo-.. . .
, .
- ~ ' ' : -, , _lL~ .
transducers in each color-sensi~ive camera receive ligh'c o~ dl~erent wavelengths from an identical narrow strip on the fruit surface.

4. Camera and Signal Formakter Circuitry The camera and signal formatter circuitry 32 (FIG. 5) of the present invenkion sequentlally reads voltage signals generaked by the photodiode arrays 63 of the segmental cameras 31, and by the red and infrared phototransducers 6~ and 5~ of the color-sensitive cameras 33. The circuitry 32 interleaves the successive readings into a serial data stream and converts each reading fro~ analog ~orm into a serial 8-bit digital word. The successîve serial wordsj in turn, are trans-mitted to a remote control room where the words are 15 demultiplexed by the demultiplexer 34; and fed to the blemish and color de'cection circuikry 35 and 39s and thence to the computer 40~ ~hlch analyzes the data to determine the proper grade category ~or each fruit.
More particularlyg the voltage signals generated in the photodlodes of the photodlode array 63 of each segmental camera 31 are read out serlally and trans-mitted to a diode array multiplexer 85. ~his multi-plexer 85J in turn, interleaves (or multiplexes) these signals with those from the other segmental cameras7 to form a compo~ite photodiode scan signal. After all the separate photodiode signals have been read out and inte~
leaved'wlth each otherg the process is repeated, cycli~
cly.
The voltage signals generated in the red and infrared phototransducers 65 and 67 of each color-sens-tive camera 33 are similarly multiplexed ~n a color camera multiplexer 87. This multiplexer 87 separately interleaves the respective red signals together to f'orm a composite red signalg and the respect~ve infrared signals together to form a composite inf~rared signal.
A~ter all of the signals have been inkerleaved in this manner the process is repeated cycllcly. Successive values o~ the CGmpoSite red signals are divided by khe corresponding infrared s~gnals in an analog divider ~9 . .

.

, to form a succession o~ color ratios.
The composite photodiode scan signal ~rom the diode array rnultiplexer 85~ along with the composite in~rared signal ~rom the color camera multiplexer 87 and the successive color ratios from the analog divider 89 are all input to a camera multiplexer 91, which interloaves these lnputs into a single analog data signal.
This analog data signal is then fed to an analog-to digital converter 93~ which converts the successive 10 analog readings into serial 8-bit binary words, and a line driver 97 then transmits the serial words to the remote control room. Control of the timing for the multiplexing operations per~ormed b~r the camera and 5ignal formatter circuitr~r is provided by a timing unit 15 A 95~ which will shorkl~ be discussed in detail.
4.1 ~
As already brie~ly descri~ed~ khe segmental cameras 31 generate analog voltage signals that are used to determine the degree Or blemish on the surface of the successive ~ruit being examined. In the presently preferred embodiment of the invention; the photodiode array 63 o~ each segmental camera 31 comprlses a linear arran~,ement of sixty-four contiguous llght-sens~tive diodes. The axls of the array is located ln the blemish 25 examining plane 499 substantially perpendicular ko the direction of travel of the ~ruitS and substantially perpendicular to a radial line from the center of the first examining region 47~ Each of the diodes gener-ates a voltage signal directly proportional to the intensity of light incident on itJ so that at any given instantj the diode array registers sixty-four separate measurements of light received ~rom contiguous sectors of the ble~ish examlning plane.
Since the four segmental cameras 31 are spaced 35 circum~erentiall~J around the blemish examining plane 49, the photodiodes generate signals representakive of light received from segments forming a 360 swakh on the sur-face of a ~rult passing through the planeO It will be appreciated that when the fruit does not completely ,~

`~

5~

~111 the ~leld of vlew o~ each ~egmen~al camera 31, some of the photodiodes (l.e.; those near the ends of each array 63j will still be examining the black back-ground area of the examining region 477 and will there-fore generate a negligible output voltage.
The timing unik A 95, as already mentioned~controls the timing of multiplexing operations performed by the camera and signal formatter circuitry of FIG. 5.
More specifically~ the kiming unik A 95 provides unique scannsr start pulses on lines 99a-99d, and a sample clock signal on llne 101 to each of the four photodiode arrays 63 in the segmental cameras 31. The occurrence o~ a scanner start pulse enables khe sample clock signal to clock out the sixty-~our analog diode voltages, there-by formlng a serial camera scan signal. The four photo-diode arrays are read in a sequentlal fashion, wlth each array receiving its particular scanner start pulse only after all sixty-four diodes in the previously accessed array have been serially read out. The four camera scan signals are transmitted to the diode array multiplexer 85 over lines 103a-103d~ respectivelyO
The diode array multiplexer 85~ shown in FIG. 5 receives the four camera scan signals on lines 103a-103dJ and time-dlvision multiplexes them together; ~o 25 generate a composite scan si~nal on line 105. Camera select signals A and B, received from the timing unit A 95 on lines 107 and 109, r~spectivelyg contr~l a sequential selection of the camera scan signals from the four segmel1tal cameras 31. The selection corres-ponds with the timing of the readouts of the respectivephotodiodes, so that an interleaving of the four camera scan signals supplied over llnes 103a through 103d is achieved.
As shown in ~IGo 5g the diode array multiplexer 85 lncludes a selectable input operational amplifier 113g such as No. HA2405~ manufactured by Harris Semiconductor of Melbourne~ Florida. Variable resistors 115 are pro-vided ak the four signal ~nputs of ~he amplifier so khat manual compensation ~or any subskantial photo-, , transducer voltage o~fsets ~an be accomplished It will be appreciated that a portion of the composite scan signal5 comprising one complete sequen-tial selection from each of the four camera scan signals, is a representation of the light lntensity received from a narrow 360 swath around a fruit ln the examining region 47. During the time elapsed while each 360 scan portion of the composite scan signal is being generated, the frult will have dropped an incremental distance through the examining region and the respectlve photo-diodes will view dlfferent portions of the frult sur- -face. Repeating the selectlon process performed by the diode array multiplexer 85~ theng results ln ~urther 360 swaths, whereby a helical-type scan of the fruit sur~ace is achieved. The clock rate is selected so that successlve 360 swaths are substantially contiguous to each other. Any changes in the velocity of the fruit as it moves throu~h the examinlng region do not affect the relative spaclng of successive swaths by a signl-~ ficant amount.
As used hereinafter the expression "camera scan"relates to the sequential data included in one readout of the photodiode arraJ 53 of one segmental camera 31.
Further~ the expression ! 360 scan relates to the sequential data lncluded in four successive camera scans~ one by each of the segmental cameras.
FIG 8 shows the composite views of each of the four segmental cameras 31 as a fruit drops ~rom top to bottom through the blemish examining plane 49. The arrangement of substantially contiguous swaths ln each view represents the sequence of scans performed by the photodiode array as the fruit drops through its field of vlew. It will be appreciated that~ since the fruit is moving while each scan is occurring the scan swaths are sloped to a slight degree.
FIG. 9 shows the images that a blemish 111 will provide when it is located approximately midway between the centers of the fields o~ view of ad~acent segmental cameras 31~ Because of the curvature of the fruit and . . --~ , .. . . . . . . .

~i because of the oblique angle at which khe blemish is viewedS it appears to be smaller than ~ts actual slze.
However, any error introduced by this viewing angle is substantially compensated for by the fact that the blemish is viewed and detected by two adjacent seg-mental cameras A sur~ace blemish is typically characterized by reflectlon of light to a substantially different degree from that assoclated with reflection from the sur-lO roundlng unblemished port~on. It is this abrupt changein reflectance at the blemish edges that the preferred embodiment of the lnvention is particularly adapted to detect and measure.
FIG. lO depicts in more detail portions of the 15 camera scan signal from one segmental camera 31 over seven consecutive camera scans. The signal is super-imposed on the outline of a blemished portion of the fruit to which it corresponds. The numbered column-like regions in the figure correspond to a sequence of 20 photodiodes~ and the signal waveforms labeled Sl-S7 represent the voltage levels of the signals for the seven consecutive camera scans. It can be readily seen from FIG. lO that the camera scan signal rlses to rela-tively high voltage levels f'or unblemished segments of 25 the fruit, and ~alls to relativel~7 low levels for blem-ished segments and for the black background area of the examining region 47. Further~ i~ is apparent that the signal voltage level tends to be lower ~'or segments near the ~ruit edge because of the obl~que angle at which 30 such segments are viewed.
The manner in which the segmental camera scan ; signals are further processed will be explained after a description of initial processing of the color-sensitive camera data.
4 2 Color-Sensltive Camera Data The color-sensitive cameras 33 generate analog voltage signals that are employed to determine the sur-face color of the successive fruit be~ng examined. Each color-sensitive camera 33 views a narrow strip of the ``;.'. , ,,.:,:. . . :
,:, ~: :
.

sur~ace of a fruit in the examining region 47, the ~trip being substantiall~7 perpendicular to the direction of travel of the fruit~ The strip is defined by the re-spective red and infrared phototransducer apertures 81 and 82~ and the light received from this strip ls focused through the respective red and lnfrared filters 69 and 71 and onto the corresponding red phototransducer ~5 or in~rared phototransducer 67, as shown dlagrammatically in FIG. 3A. me phototransducers then generate signals 10 at voltages proportional to the intensities of the light they receive.
As shown in FIG. 5, the voltage outputs of the various red and infrared phototransducers 65 and 67 are suitably buffered in buffers 117; then the "red i! signals a~e transmltted over lines ll9a-119a, and the "infrared !~
signals transmitted over lines 120a-120d, all to the color camera multiplexer 87. The camera select signals A and ~, recelved on lines 107 and lO9g are used to select sequentially from the various buffered photo-transducer outputs, whereby a composite red signal anda composite infrared signal ~re generated. In a fashion similar to the generation of the composite scan signal by the diode array multiplexer 85~ the sequential read-ing o~ the phototransducer voltages~ coupled with the movement o~ the frult through the examining reglon 47, re~ults ln a helical-type scan of the ~ruit sur~ace.
The color camera multiplexer 87 includes two selectable input operational ampli~iers 121, one for generating the composite red signal and the other ~or 30 generating the composite infrared signal. Variable resistor~ 123 are provided at the inputs of the ampli-fiers so that manual compensation for any substantial phototransducer voltage offsets can be accompl~shed.
The composite red and in~rared signals are out-35 put ~rom the operational amplifiers 121 on lines 125 and127, respectivelyy and transmltted to the analog divider circuit 89, which generates~ in real time9 the ratio of the magnitude o~ the red signal to that o~ the infrared signal. The analog div~der 89 can beg for example, .. .. .. . .
. ~
:~ , -~ .

Part No. BB4291~ manufactured by Burr-Brown Re~earch Corporation o~ Tucson9 Arizona. The divider 89 includes an integral low-pass ~ilter 129 in its output stage, for eliminating spurious voltages that might occur at the transitlons between successive red and in~rared read-ings. It will be appreciated that the color ratio signal generated by the divider comprises a sequential representation of tle ratio of red light intensity to infrared light intens-lty for a succession of ~ruit surface portions forming a helix on the sur~ace of the fruit.
The color ratios generated in the aforedescribed manner are substantially insensitive to variations in illumination intensiky and to variations in the pro-portion o~ the fields ~ view of the color cameras 33that is occupied b~ the ~ruit. Any such variations would result in corresponding variations in both the red and infrared phototransducer measurements, and thus would be substantially self-cancelling in the ratio cOmputations~
4.3 Composite Camera Data The composite segmental camera scan signal on line 105 khe composite infrared signal on line 127~ and the color ratio signal on line 131 are all transmltted to the camera multiplexer 91J which interleaves the three slgnals to form a combined analog daka slgnal on line 133. Data select signals C and D, supplied over llnes 135 and 137 ~rom the timing unit A 95, control the lnterleaving by deletlng the first and last photo-3 diode readings in the sequence of sixty-four readings in each camera scan of the compos~te segmental camera signal, and inserting in their respective places the color ratio signal derived from the corresponding color-sensitive camera 33~ and the infrared color signal de-rived from the color-sens~tive camera 33 next in se-quence.
Thus~ the analog data signal on line 133 com-prisesg in sequence the infrared color signal and color rat1O signal derived from one color-sensitive camera 33;
.

- . . .
- , ~ , , .
, .
.

.

~ollowed by 62 readings derived from the corresponding segmental camera 31. This is followed~ ln turng by the same sequence o~ signals derived fro~ the next asso-ciated pair of cameras.
As will be explained in more detailg the succes-slve readings of the segmental cameras 31 are used by the blemish detection circuitry 35 (FIG. 3B)g to obtain a measure of blemish on the surface of each fruit. The infrared color signal and the color ratio signal will both be used by the color detection clrcuitry 39 (FIG.
3~). The infrared color slgnal will be used to deter-mine whether or not a portion of a frult surface ls being examined and the color ratlo signal will be used to obtain a measure o~ the color of that fruit sur~ace port~on.
The deletion of two photodiode readings from each sequence o~ sixty~four does not significantly affect the blemish detection capability of the invention appar-atusj because the remaining sixty-two readings can adequately cover a fruit in the blemish examining plane 49. Moreover, the signals derived from the firsk and last photodiodes on present commerc~ally available photodiode arrays are generally less reliable than those derived from the other photodiodes.
The analog data signal on line 13~ is transmltted to the analog-to-dlgital converter 935 ~or conversion to a correspondlng digital data ~ignal. The converter 93 can be for example, an ADC 82, manufactured by Burr-Brown~ and it provides a serial output comprising a sequence o~ 8-bit words~ In addition, an end-of-con-;~ version pulse is generated at the end of each such 8-bit segment. An A/D clock signal on line 139 from the timing unit A 95 controls the conversion performed by the analog-to-digital converter 93. The clock signal com-prises sequential bursts o~ eight clock pulsesS one such burst occurring for each independent reading in the an-alog data sign:al. The analog-to~digital converslon is performed primarily to facllitate transmission of the data more easily over a lengthy cable to a remote control ,~

.
~, . ..
. . . , .
: ' : .

-22~
roomg where the remaining equipment of the system can be better protected from the en~ironment of the ~ruik transport structure.
The digital data signal and khe end-of-conversion signal are transm~tted over llnes 141 and 143, respec-tively, to the differential line driver circult 97, which, in kurng transmits the two signals on cables 145 and 147, respectivel~. Addltionally, the timing un-lt A 95 transmits a clock slgnal and a scan sync signal on 10 lines 149 and 151 respectively, to the dlf~erential llne driver circult 97, which, in turn9 transmits these two signals on cables 153 and 155s respectlvely. The cables 145, 147~ 153 and 155 are routed to the remote control room where the demu~tiplexer 34 and the blemish 15 and color detection circuitry 35 and 39g respectively, are located.
Also routed to the remote control room ls a re-set timing signal on line 159 generated by the timing unlk A 95, in response to receipt of periodic reset pulses on line 161 from a sensor (nok shown) ad~acent to the first conveyor 230 The sensor generakes a pulse on detection of a conveyor tray 41 on which a ~ruik ~s carrled. The timing unit A 95 includes adjustable delay means for allowing manual adjustment of a time delay between the receipt o~ each reset pulse on line 161 and the generation of a pulse in the reset timing slgnal on line 159.
This completes the descripkion of the generation9 multiplexing and formatking of signals derived from the cameras 31 and 33. Accordingly~ the following descrip-tive sections~deal with demultiplexing and utilization of the signals.
5. Demultiplexer The demultiplexer 34 (FIG. 3~ separates the successive serial 8-bik binary words received ~rom khe camera and signal formakter circuitry 32 inko separate sequences o~ blemish words~ color ratlo words and in-~rared words~ Each blemish word corresponds to a read-ing of one photodiode in the photodiode array 63 of one . ~

eegmental camera 31. Each infrared word corresponds to a reading of the infrared phototransducer 67 of one color-sensitive camera 33s and similarl~3 each color ratio word corresponds to a ratio of readings of the red and infrared phototransducers 65 and 67 from one color-sensitive camera 33. The blemish words, color ratlo words, and in~rared words are subsequently pro~
cessed in the blemish detection circuitry 35 and color detection clrcuitry 39.
As shown in more detail in FIG. ~, the digital data signal on cable 145~ the end-of-conversion signal on cable 147 the clock signal on cable 153 and the scan sync signal on cable 155 are received by a conven-tional differential line receiver circuit 163$ which reconverts the signals to "single-ended" logic. The differential line receiver circuit 163 comprises four separate line receivers, such as Part No. SN 75115g manufactured by Texas Instruments, Inc. J of Dallasg Texas, along with appropriate reslstor terminations to match the characteristic impedance o~ the cables.
: A tlming unit B 171 receives the end-of~conver-sion s~gnalg khe bit clock signal and the scan sync signal over lines 1659 167 and 169g respectively, f'rom the line receiver circuit 153. The timing unit ~ 171 also receive~ the reset signal directly over llne 159 ~rom tiilng unit A 95~ and generates all the timing signals required by the demultiplexer 157g the blemish detection circultry 35 and the color detection circuitry 39.
3 The digital data signal and the bit clock signal are transmitted over lines 173 and 165 ~rom the line receiver circuit 1O3 to the demultiplexer 34. The demultiplexer 34, as shown in more detail in FIG. ~g converts the digital data ~rom a serial format to a parallel rormat and demuItiplexes the various digitized components o~ the composite signal, ~.e., the sequential measurements of the composite scan signalg the readings of the composite infrared color signalg and the com-puted ratios o~ the color ratio signal. Serial-to-. . , .

' parallel conversion is performed by a conventional 8 bit shif-t register 175 into which the digital data signal is clocked by the clock signal on line 165. The eight bits stored in the shift register 175 at any given time, are registered on lines 176 from its eight output ter-minals.
A blemish word clock signal, a color word clock signal, and an infrared word clock signal, all supplied from the timing unit B 171 on lines 177, 179 and 181, respectively, control the demultiplexing function of the demultiplexer 34. The color word clock signal on line 179 is utilized to clock the eight-bit output from the shift register 175 into a color ratio word latch 183, and comprises a sequence of pulses each occurring in the first blemish word period in each camera scan, when the 8-bit word corresponds to a color ratio word. Simi-larly, the infrared word clock signal is utilized to clock the eight-bit output from the shift register 175 into an infrared word latch 184, and also comprises a sequence of pulses, each occurring in the sixty-fourth blemish word period in each camera scan, when the 8-bit word corresponds to an infrared word.
The blemish word clock signal is utilized to clock the eight-bit output from the shift register 175 into a blemish word latch 182, and comprises a sequence of pulses, each occurring whenever the eight bits then stored in the shift register correspond to either a blemish word, a color ratio word, or an infrared word.
Color ratio and infrared words are inhibited from being clocked into the blemish word latch, however, by an in-hibit signal supplied on line 188 from an OR gate 188a, which OR's together the color word clock signal and the infrared word clock signal, received on lines 179 and 181~ respectively.
At the end of each word time, the word is clocked into either the blemish word latch 182, the color ratio word latch 183 or the inrared word latch 184, as ap-propriate. The blemish word latch 182 outputs a blemish word sequence signal on lines 185, the color ratio latch 183 outputs a color ratio word sequence on lines 186 and the ln~rared latch 184 outputs an in~rared word sequence signal on lines 187.

6. Blemish Detection Circuitry The blemish detection circuitry 35~ shown in de-tail in FIG. 7, receives the successive demultiplexed blemish words on lines 185 from the demultiplexer 34, and analyzes the words to determine the total amount o~
blemish on the surfaces of the successive fruit being examined. For each segment of a fruit belng examinedg its corresponding blemish word is compared to blemish words ~or neighboring segments~ to obtain a measure of change in reflectivity ~or that portion of the fruit surface. In the presently preferred embodiment of the invention, the comparison is made by dividing each blemish word b~J the average o~ either the two immedi-ately preceding blemish words er the two immediately subsequent blemish words for the corresponding photo-diode.
The successive blemish word comparisons are per-formed by a scan storage register 189~ which stores blemlsh words corresponding to the two ~mmed$ately pre-ceding 360 scansJ a scan select c~rcult l91, which formats the data into successive numerators and denom-inators, and a digital divider 193, which per~orms the act~al d-lvision. The successive blemish word quotients9 generated b~ the digItal divider 193~ are ~iltered in a digital h~gh-pass .~ilter 195 to remove any slowly vary-ing elements that might be presentg such as those intro~- -duced by the curvature o~ the ~ruikO
A digltal blemish integrator l99 then integrates the successive ~iltered words derived by the dlgital high-pass filter 195, to obtain a measure of total sur-~ace blemish ~or each fruit~ A blemish on/of~ timing circuit 197 controls the integrator l99 so that only words correspondin~ to actual segments of the frult sur~ace~ as contrasted with the black background of the examining region 47, are integrated.
.l Scan Stora~e Register . :.. . : '' s~
-26~
~ he scan storage register 189 comprises a pair of 8 x 256 bit shift registers for storin~; the parallel 8-bit blemish words for two successive 360 scans b~r the four segmental cameras 31. The blemish word se-5 quence signalg which contalns the successive 8-b~;t blemish words, is received on lines 185 from the demul-tiplexer 32~ and the successive words it contains are clocked into the scan storage register by the blemish word clock signal on llne 177.
The scan stora~;e register 189 provides two par-allel 8-bit outputs~ the first output being on lines 203 and comprising the blemish word sequence signal delayed by 256 blemish word times (i.e., delayed by one 360 scan by the ~our segmental cameras 31)9 and the second 15 output being on lines 205 and comprising the blemish word sequence signal delayed by 512 blemish word times (i.e.g delayed by two 360 scans by the four segmental cameras). Thus~ at any given time, the blemish word sequence signal on lines 185 and the scan storage regis-20 ter's first and second outputs on lines 203 and 2057respectively; contain blemish words corresponding to the same photodiode for three consecutive 360 s¢ans.
5,2 Scan Select Circuit Successive comparlsons Or blernlsh data words are 25 accomplished b~ successively digitally dividing each blemLsh word by one half khe sum (i.e., the average) of the two blemish words corresponding to the same photo-diode for either the two immediately preceding scans or the two immediately subsequent scans. ~3ach resultant 30 quotient is a measure o~ the percentage rate of change of re~lectance for the corresponding portion of the surface of the fruit being examined.
A substantially identical measure of the percen-tage rate o~ change of' surface reflectance could be 35 accomplished by successively dividing the blemish words corresponding to ad~acent photodlodes wlthin each scan.
Typical photodiode arrays that are presently available commercially, however~ suf~er the drawback OL' having small voltage offsets between ad,jacent photodiodes.

Such offsets would produce errors in the quotients gen-erated by the divislon operation. In ~he pre~erred embodiment clescribed aboveg on the other hand~ where the divlsion operation is performed with blemish words corresponding to the same photodiode onlyg these voltage offsets are substantially cancelled.
The scan select circuit 191, shown in detail ln FIG. 119 formats the successive blemish words into appropriate numerators and denomlnators for processing 10 by the digital divider 193. As shown in FIG~ 7, each parallel 8-blt blemish word that is received by the scan storage register 189 on lines 185 is also transmitted to the scan select circuit 191 Simulkaneouslyg the words corresponding to the same photodiode for the 15 previous two scans are transmitted over lines 203 and 2059 respectivel~Tg to the scan seleck circuit. Accor-d~nglyg this circuit 191 receives the three parallel blemish words, and provides an appropriate sequence of numerators and denominators to the digital divider 193.
It is desirable that the digital divider 193 should never divide by a number near zeroS i.e., by a blemish word having eight successive zeros, as would result if a photodiode had no light incident on it.
Dividing by a number near ze.ro creates a likellhood that the quot-lent will exceed the limits of the divider and th~t an erroneous output will result. At those tim2s when a ~ruit is ~ust enterin~ the f~elds of view ~ the photodiode arrays 63, the current blemish words will likely be non-zeroy while those for the preceding two scans9 which correspond to the black background area o~
the examin~ng region, will be at or near zero. Thus, if the digital divider 193 were to divlde the blemish words of the current scan b~y the average of those of the preceding t~o scans, erroneous output quotients could be generatedO
To alleviate this problemg the ScQn select cir-cuit 191 insures that the successive denominators pro-vided to the digital divider 193 never correspond to the black background area. When the first half o~ a fruit ,, .

is being exa~ined the numerators are forrned by the successive blemlsh words from the second preceding 360 scan9 and the denominators are formed by the ~verages of the success~ve blemish words from the current 360 5 scan and the immediately preceding 360 scan. On the other hand~ when the last half of the ~ruit is belng examinedS the numerators ~re formed by the successive blemish words from the current 360 scan3 and the denom-inators are formed by the averages of the successlve blemish words of the preceding two 360 scans.
In this manner, whenever any portion of the fruit is being examined~ the denominator provided to the divider 193 will always be based on blemish words cor-responding to segments located furthest from an edge of 15 the ~rult. Accordinglyg the scan select circuit 191 -minimizes the likelihood of having a denominator near zero3 and thus of having erroneous output quotients from the divider 193. Each such quotient, theng is an accur-ate measure of the rate of change of surface reflectance for a particular portion Or the frultO
As shown in FIGS. 7 and 11 the scan select cir-cuit 191 rsceives the blemlsh word sequence signal on lines 185 ~rom the demultiplexer 32j and re¢eives the ~equences of blemish words for the immediately preceding 360 scan and the second preceding 360 scan on lines 203 and 205, respectivelyg ~rom the scan ~torage regis-ter 189. For each camera scan, the scan select circuit makes a word-by-word comparison of blemish words from the current 360 scan with blemish words from the second preceding 360 scan9 detecting which o~ the two scans is first to include a blemish word corresponding to a segment of the fruit~surface3 as contrasted with a portion of the black background area.
This comparison is accomplished using ~irstg second and third OR gates 207, 209 and 211g respectively, and first and second D-type flip-flops 213 and 2153 respectively. The four most significant bits in the blemish words of the current scan are successively OR'ed in the first OR gate 207, and similarly~ the four most , signlficant bits for the words of the second precedlng scan are OR'ed in khe second OR gate 209 . It will be appreciated that the output o~ OP~ gates 207 and 209 on lines 208 and 210g respectively, are "frult present' signals which are a logical "1" whenever the corres-ponding blemish words correspond to segments o~ the sur~
face o~ the fruit being examined. These signals on lines 208 and 210 are applied as lnputs to OR gate 2119 the output of which is connected to the D input terrninal of ~lip-flop 213.
As soon as the output of either of the OR gates 207 or 209 goes to a logical "1", a logical "1" is clocked lnto the first flip-flop 213 b~ the blemish word clock signal on line 177. The Q output of the first flip-flop 213) in turn, clocks the output of the second OR gate 209 into the second flip-flop 215. Thusg if the particular camera scan from khe second preceding 360 scan was the first to contain a word corresponding to a segment of the fru~t, then the second half of the fruit is being examined and the Q output of the second flip-flop 215 ls a lo~;ical "1". On the other hand, if the present camera scan is first to contlnue a word corresponding to a segment of rruit~ the ~lrst half o~
the ~ruit is being examined and the Q output o~ the second fllp~~lop 215 ~s a logical "O'. The process is repeated for each camera scan.
In accordance with the outcome of the above comparison, the scan select clrcult 191 generates) succe~sivel~J~ the appropriate numerators and denomina-3 tors to be provided to the digital divider 193 on lines217 and 219~ respectively. This is accomplished uæing ~irst and second digital data selectors 221 and 223 and a digital adder 225. Each of the data selectors 221 and 223 comprises a pair of quadruple 2-line to l-line data selector multiplexer~.g such as Part No. 7~ IS 157 ~anufactured b~J Texas Inskruments of Dallas, Texas.
Each of the data selectors 221 and 223 receives two parallel 8-bit data inpuks~ one being the successive blemi.sh words for the current 360 scan on lines 185 .
.: .

.

-3o-from the demultiplexer 34, and the other be~n~ the successlve blem-Lsh words for the second preceding 360 scan, on lines 205 ~rom the scan storage re~ister 189.
me Q output of the second flip~flop 215 ls provided on line 227 to the SELECT input of the first data selector 221, while the corresponding ~ output is provided on line 229 to the SELECT inpu~ of the second data selector ~33.
If the Q output of the second flip-flop 215 i3 a loglcal "1" (and the Q output a logical "zero"), then the first data selector 221 automatically selects the blemlsh word data for the current 360 scan and outputs such par~llel data on its output terminals3 and the ~econd data selector ~23 automati~ally selects the 15 blemish word data for the second preceding 360 scan and outputs such parallel data on ~ts ~tput terminals.
On the other hand, if the Q output of the second flip-flop is a logical l'zero" (and the Q outpuk a logical "1"), then the first data selector outputs the blemish 20 word data for the second preceding 360 scan; and the second data selector outputs the blemish word data for the current 360 scan.
The output of the second data selector 223 is transmltted over l~nes 231 ko a first set of lnput termlnal~ on the cligital adder 225, whlle the succe~-~ive blemish words for the immediately precedi~g 360 scan are transmltted over llnes 203 from the storage register 189 to a second set of input te~ninals on the adder. The adder arithmetically surns the two parallel 8-bit inputs, providing a parallel 8-bit data output and a ~ARRY output. The seven most signi~icant bits of the data output in comblnation with the CA M Y outputg con-~titute a sequence of 8-bit words, each of which ls one half the sum (i~e., the average) o~ the corresponding two 8-blt blemish words received by fne adder. It will be appreciated that use o~ the CARRY output and the seven ~ost significant bit~ of the sum is effectively shiftin~ the sum one bit to tl~e right, ~lh~ch is a divlde-b~J-two operation.

,. . . . . . .
., ~ '' , ' :

~ ~ ~ 5~

The output of the fir~t data selector 221 on llnes 217 forms the successive numerators for proaessing by ~he digital d:Lvider 193. The seven most significant output bitsg along with the CARRY output, of the adder 225, on lines 219, form the successive denominators for processing by the divider 193 6.3 Di~ital Divlder The dlgital divider 193 divides each o~ the successive numerators received on lines 217 by khe cor-responding denominators received on lines 219g to obtaina sequence of quot~ents that measure the rate of change o~ reflectivity of the surface of the fruit belng examined. The blemish word clock signal on llne 177 is u~ed by the divider 193 to control its sequence o~ oper-15 ation. The dlv~der output is a parallel 9-bit quotient ssquence signal on lines 233.
The quotient sequence signal comprises nine parallel bitsg with the most signi~icant bit repre-senting 21, and the least signiflcant bit representing 2-7. Slnce the quotient is norrnally about l~Og and at the fruit edges, less than l~Og the divider capacity of 3.99 ~s rarely exceeded. The digltal divider 193 can be readily constructed using conventional design technlques descri~ed ln many handbooks on digital circuit design, such as Fairchild TTL ~pplicatlons Handbook, published by Fairchlld Carnera and ~nstrurnent Corporation of Moun-tain Vlew, Call~ornia~ 19'73.
6.4 High Pass ~ilter The digital high pass filter l9~g shown in detail in FIGS. 12 and 13 receives the quotient sequence signal on lines 233 and substantlally eliminates the constant and slowly varying portlons of the signal~
particularly those caused by the curvature of the fruit ~urface being examined. The illustrative filter com-prises a pair of identical cascaded one-pole filter sectionsg FIG. 13 showing one such section. Conven-tional two's complement binary coding ls usedJ so that negative numbers can be conveniently handled. The ~ilter sections provide an output comprising el~ht .' " ' ' S~
~32-parallel bits of magnitude data and one bit of sign datag khe latter indicating whether the magn~tude is positive or negative. These filter sections can also be implemented using conventlonal digital circultry techniquesg such as descrlbed in the aforementioned Fairchild TTL Applications Handbook.
It will be appreciated that many high-pass filter designs can be used to achieve the goal o~ elimlnating constant and slo~ly varying portlons of the quotient 10 sequence signal. The presentl~J preferred filter designg provides suf~icient ~iltering to substantially eliminate the undesired portions of the input signal, yet it can be readily implemented without undue circuit complexity.
Following the two cascaded filter sections -ln 15 the high-pass filter 195, is an absolute value stage 239 for converting the negative portions of the ~iltered signal into positive portions o~ a corresponding mag-nitude. In this manner, the detectiGn of a rapid de crease in sur~ace re~lectivity is a~forded the same 20 weight as the detection o~ an equally rapid increase in sur~ace reflectivity. The output terminals o~ the absolute value stage 239 ~orm the high-pass ~ilter output signal on lines 245.
The absolute value stage 239 comprlses a pa-lr of quad 2-inpuk exclus~ve OR gates. The eight parallel blks o~ magnitude data from the fllter sections are supplied lndividually to one set of inputs on the eight gates, whlle the sign blt ~rom the filter sections is supplied to all eight o~ the second sa~ of inputs. In this manner~ i~ the sign bit is a "æero" (lndicating a posltive magnitude~ then the outputs of the ei~ht exclusive-OR gates will correspond to the eight parallel - bits of magnitude data from the filter sections. On the other handJ i~ tlle sign bit is a ~1 ?1 ( indicaking a 35 negative magnitude) then the outputs of the ei~ht ex-clusive-OR gates will correspond to the complement (i.e.g the inverse9 in two's complement binary coding) o~ the eight parallel bits o~ magnitude data ~rom the filter sections.

... . .
. . ' , , : ', ;' ': ' , . ' ' . "

6.5 Rlemish On,/O~ T~mln~ Circuit The blemlsh on/off timing clrcuit 197 ~'IG. 7) generates a blemish timing slgnal on line 247~ whlch enables the blemish integrator 199 to sum the succes-~ive filtered d,gital quotients supplied on lin~s 24~from the high-pass filter, to obtain a measure of total blemish on the surface of each fruit being exam~ned.
The blemlsh timing signal is a logical "l", thereby allowing the integrator 199 to operate, only when seg-10 ments of the fruit sur~ace, as contrasked with segmentsof the black background area of the examining region 47J are being examined.
The blemish timing signal remalns ln the logical "zero" state, however; when segments of the fruit ~,ur-15 face at or near the edges of each fruit image~ arebeing examined. Because such segments are vlewed at oblique angles~ and the corresponding blemish words are not completely accurate measures of the reflectivlty of the fruit surface, it is desirable to treat such seg-ments near the fruit edges in the same manner as thebackground area. ~here is sufficient overlap in the portions of the fruit surf~ce viewed by each segmental camera 31 that the elimination of three blemish words correspondlng to the frult edges in each camer.a scan, 25 13 not slgnificant. All or nearly all of the portions of the fruit surface corresponding to eliminated blemish words, are also viewed by an adJacent segmental camera 31, and are not normally eliminated from the camera scan ~or that camera.
The blemish tlming signal on line 247 ls genera-ted by detect~ng, for each camera scan, the i~age t'envelope' of a fruit being examined (i.e., the timing of the blemish words corresponding to segments of the surfaGe of the fruit, as contrasted with the black background area), and by then eliminating three blemish word times from both the leading an,d trailing edges of the envelope. Add~tionally, the blemish on/off timing circult 197 includes circuit means for differentiating between a blemish and the trailing edge of a fruit .

~s~

image envelopeg so that the blemi~h timlng signal on line 247 remains in the logical "l'i state even when a non-reflective surface blemish is being examined.
Thusg the blemish integrator 199 remains enabled to sum the successive blemish quotients on lines 245g until the actual trailing edge of the fruit im~æe is reached.
The blemish timing signal on line 247 is gener-ated using the "fruit present" signals on lines 208 10 and 210, received from the scan select circuit 191.
It will be recalled that the ~ruit present ~lgnal ls ln the loglcal "1' state only when the corresponding blemish word corresponds to a segment of a rrult sur-face, as contrasted with the black background area.
15 The fruit present slgnal on line 208 corresponds to the current 360 scan9 while the slgnal on line 210 corres-ponds to the second preceding 360 scan. The ~ccur-rences of non-reflective blemishes~ however~ cause the fruit present signals to have "dropouts ~IJ Just as though 20 t~ne trailing edge o~ the fruit had been reached and the black background area was being examined. The blemish timing signal corresponds to the fruit present signal, but with the dropouts due to blemlshes removed and with three blemish word periods deleted from all leading ancl 25 trailing edges o~ the ~ruit image in each camera scan.
As shown in detail in FIG. 14, the blemish on/off timing clrcuit 197 comprises ~irst and second OR gates 251 and 253J respect-lvely, and AND gate 255, a 12-bit counter 257, a 250-bit shift register 259 and a 6-bit 30 counter 261.
The circuit 197 lnitially generates~ for each successive camera scan, a partial envelope signal on line 263, which defines an envelope of the fruit image but with slx blemish word periods deleted from both its 35 leading and trailing edges. This partial envelope signal is generated in a recursive fashion, by succes-sively OR'ing in the first OR gate 251 the ~ruit present signal for the current scan~ received on line 2089 with the partial envelope signal for the corresponding camera ~, ~

scan of the previous 360 scan (iOe., the prior scan for the same segmental camera 31). Thusg the output o~
the OR gate 251 is a logical "lli whenever the fruit present signal is a loglcal "1 9 and is held in that 5 state by the partial envelope signal even if a non-re~lective blemish causes a dropout in the fruit present slgnal.
The output of the OR gate 251 is connected to the ENABLE input of the 12-bit counter 257, which for each camera scan deletes the first twelve blemish word periods of logical "1" state from the OR gate Z51 out-put. The counter 257 produces zero-state outputs so long as its ENABIE input is zero~ and continues to produce a zero output for the first twelve l's applied 15 to its ~NABLE input, a~ter which the output signal ~ollows t`ne ENAELE input signal. The counter 257 is reset between successive camera scans b~J a reset signal on line 265 from the timing unit ~ 171. The output of the coun~er 257 is connected to the shift regi3ter 259 which delays the ou~pu~ by 250 blemish word periods, to produce the partial envelope signal on line 263. It will be appreciated that the dela~J of 250 blemish word periods effectively shi~ts the envelope slgnal out of phase by six periods~ since there are 256 periods in a complete 360 scan. The envelope on line 263 there~
~ore has lts leading and trailing edges shortened by ,six periods.
The partial envelope signal on line 263, in addltion to being connected to one input terminal of the first OR gate 251 ~o form the partial envelope signal for the next 360 scan, is connected to one input terminal of the second OR ~ate 253. Colmected to the second input kerminal of the OR gate 253 i~3 the output of the AND gate 255J which ANDs the two fruit present signals (present scan and second prevlous scan) received on llnes 208 and 2IOj and produces an output signal which is the shorter of the two lnput envelopesg and includes dropouts due to blemlsnesO The output of the second OR gate 253, then~ represen',,s an envelope of the . ..

. ' :

shorter of 1) the frult ~mage ~or the present ca~era scan and 2) the ~ruit image for the corresponding camera scan for the second previous 360 scanj but wlth drop-outs due to non-reflective blemishes being deleted.
The output of the second OR gate 253 is connected to the ENABLE input of the 6-bit counter 261, which, for each camera scan; deletes the ~lrst six blemish word times of logical "l; from the OR gate 253 outputg thereby formin~ the blemish timing signal on line 247.
The 6-bit counter 261 ~unctions in the same way as the 12-bit counter 257. It provides a zero output when the ENABLE input is zero9 and maintains a zero output for the first six "one inputs, after which the output signal follows the input signal. This has the effect of deleting the first six "ones from the leading edge of the envelope signal. An inherent property of the high-pass filter 195 is that it delays the output by three blemish word periods. Accordingly, the phase relationship bet~eell the blemish timing slgnal on line 247 and the rilter output signal on lines 245 is such that the blem-lsh integrator 199 is disabled for the first three and the last three blemish word times of each camera scan.
6 ~ Blemish Integrator The blemish integrator 199 (FIG. 7) sums together the su¢ces~ive digltal words of the high-pass ~ilter output signal to derive a blemish count signal on lines 275 that is a measure of the total blemish on the sur-face of each fruit. The summing activity is enabled by the blemish timing signal on line 247g which is in t'ne logical "1: state only when the high-pass filter output signal contains data based on segments of the fruit surface, as contrasted with portions of the black back-ground area. The integrator 199 is reset to the logical "zero" state by a reset slgnal on line 159 from timlng unit A 95 (FIG. 5)g immediately prior to the examinat~on of each fruit.
The blemish lntegrator 199 may be implemented in any of a variety of formsO For example, it may in-, .~.

.,........................................ :~

.
..

'3~

clude an 8-bit adder having an over~low signal connec-ted to increment an up/down counterg the several stages of which supply the output signals on lines 275. As will be appreciated from the following descriptive sectiong the up/down counter may be decremented to compensate for erroneous blemish indications.
It will be appreciated that the examination o~
~ruit that appears unblemishedg will sometimes result in a non-zero blemish measurement by the blemish inte-grator 199. This is caused by the detect~on of stemand blossom ends, by the fruit sur~ace texture, and b~J
random noise in the system. It is pre~erableg howeverg that the blemish detection circultry 35 compensate for these factors and provlde a blemish measurement that is nominally zero for unblemlshed fruit. This is accom-plished by a normalizer clrcuit 295.
6 7 Mormalizer Circuit The normalizer circuit 295 (FIG. 7) generates a blemish normalizer pulse sequence on 11ne 297 that is transmitted to the blemish integrator l99g ~or decre-menting the blemish count signal on lines 275. me frequency of the pulse sequence on line 297 is manuall~J
sel0ctableg and the pulse sequence ls "ENA~,LED" by the blemish t-lmlng signal on line 2~7, i.e.g only when segments o~ the ~ruit surface are being examined. By an emplrical selection o~ the frequency of the pulse se~uenceg the blemish count signal on lines 275 ~rom the blemish integrator 199 can be made to be near zero for unblemished fru~t. Higher counts are,indicative of fruit having greaker surface blemish.
It will be understood b~; those of ordinary skill in the art that the normallzer circuit 295 can be con-structed using known design techniquesO For exampleg the normalizer 295 can be merely a blnary counter for frequency-dividing input pulses supplied from the blemish word clock on line 197, and ~or prov~ding a string o~ output pulses at a controllable rate to the blemish integrator l99g where the pulses are utilized to dlminlsh the overall blemlsh indicationg such as ~y , decrementing the UP/DOWN counter in the blemish inke-grator.
6.8 Size Detection Circuit~
The size detection circuitry 37 (FIG. 7) gener-ates a size count signal on l-lnes 311 that is a measure of the s~ze of each fruit being examined. The circuitry counts the number of segments in the total reflective surface of each fruik, by counting the number o~ blemish word periods that the blemish timing signal on line 247 is in the logical t'l'l state. The circuitry 37 is reset to the logical "zero 1I state by the reset signal on line 159, lmmediately prior to the examination of each fruit.
It will be understood by those of ordinary skill in the art that the size detection circu-ltry 37 is basically a multistage binary counterJ and that it can be readily constructed using known design techniques.

7. Color Detection Circuitry The color detection circuitry 39g as shown in detail in FIG. 15, receives the demultiplexed color word sequence signal on lines 185 and lnfrared word se~uence signal on lines 187 from t'ne demultiplexer 34 (FIG. 3B). The color detection circuitry analyzes the æuccessive words of each signal to obtain measures of the surface color of each fruitg and to obtaln an addi-tional measure of the slze of each fruit. The circuitry39 ~enerates (1) a color count signal on lines 317~
wh-lch is derived by summing together normalized color ratio words for all of the surface strips on each of the successive fruitg (2) an excess color count signal on lines 319, which is a count of the number o~ sur~ace strips on each of the successive fruit for which a selectable color level is exceededg and (3) a color size count signal on lines 321, which is a count of the number of surface strips on each of the successive ~ruit.
- As previously describedg the successive color ratio words received on lines 186 are each derived by dividing the output of a red phototransducer 55 by the output of the corresponding infrared phototransducer 67.
The magnitude of the color ratio word is a measure o~
.

s~

color or ripeness of the correspondin~ strip on ~he surface of the fruit being examined~
It is pre~erable that the color count signal on lines 317, which is generated by the color detection 5 circuitry and which is a measure of the surface color o~ each fruit; be normalized so that ~t is near zero ~or ripe fruity with greater magnitudes for green and re-greened ~ruit. Normaliz~ng is accomplished by a color normalizer circuit 323 thak successively sub-10 tracts each color ratio word received~ on lines 186,from a manually selectable reference level. The color normalizer circuit 323 is clocked one pulse for each successive color ratio word. It outputs a normalized color word sequence signal on lines 325.
The successive normalized color words on lines 325 are summed together in a color integrator 327g to generate the color count signal on lines 317. The i~-tegrator 327 is clocked b-J a clock signal on line 329.
As will be further explainedg this clock slgnal includes a pulse for each color word corresponding to a surface strip of a fruit~ as contrasted with the black back-ground area of the examining region 47.
The clocl~ sigllal on line 329 is derived from an AND gate 333 which has one input connected to the color 25 word clock on line 179 and a second input connected to the output of a comparator 331. The comparator 331 com-pares each of the successive infrared words received on llnes 187 to a suitable manually selectable threshold value, to determine whether each infraréd word~ and thus ~ts corresponding color ratio word~ correspond to a surface strip of a fruit~being examined or to the black background area. If the threshold is exceeded~ it iB
assumed that a fruit ls being examined and the output of the comparator 331 will be a logical 1'1 , thereby enabling the clock si~nal on line 329 ~rom the AND
gate 333. The clock signal on line 329 -ls therefore equlvalent in timing to t'ne color word clock signal on line 179, but is enabled only during examination of a fru-Lt, as determined in the comparator 3310 .
.
, , ,. ~

- ~o - :
An excess color comparator 337 and an excess color counter 339 generate the excess color count signal on lines 319. The excess color count indicates ~or each fruit the number o~ normalized color rati~ wordsg 5 on lines 325, l~hose ma~nitudes exceed a selectable reference thresholcl. The excess color counter 339 is also clocked by the clock signal on line 3299 whichg it will be recalledy includes a, pulse ~or every color rat-lo word corresponding to a surface strlp of' a fruit. These 10 two circults can be utilizedg for example, to determine the proportion of a fruit which ls greener than a pre-determined level.
A color size counter 341 counts the number of separate surface strips in the total reflectlve surface 15 of each fruit; to produce the color size count signal on lines 321. The counter 341 accomplishes t'nis by counting the successive clock pulses in the clock signal on line 329, which contains one pulse for every infra-red word that corre~ponds to a sur~ace strip of the 20 fruit being examined.
The color integrator 3279 the excess color counter 339 and the color size counter 341 are all reset to the logical "zero state by the reset slgnal on line 159 (omitted for clarity in FIG. 15)q immediately prior to 25 the examination of eac',l ~ruit. In th~s manner, k~e counts ~or each ~rul~ are independent of the counks Eor ~rui~ previously examlned.
It will be understood by tl~ose of ordinary skill in the electronics art that the various circuit elements 30 of' the color detectlon circuitry 399 described aboveg can be readily constructed using commercially available digital integrated circuits in accordance with known design techniques. More ~pecificallyg the circuit ele-ments comprise comparators; adders and counters. The 35 comparators 331 and 337 are conventional dlgital com-paratorsg ti~e counters 341 and~ 339 are conventional binary counters5 the normallzer 323 is a digital adder~
and the color integrator 327 is basically an accumulating adder.

, .
' . ' . ~,

8. Fruit Gradin~
As previously described~ the blemish detectlon circuitry 35 produces a dlgital blemish count of the total surface blemish o~ each of' the fruit being ex-amined~ and a dig~tal size count signal on llnes 3119which is a successive count of roughl~ the number of surface segmen'cs of each fruit. Simultaneouslyg the color detect~on circuitr~J 39 produces a digital color count signal on lines 317g which is a successive summa-tion of all the color ratio words for each fruit, adigital excess color count signal on lines 3199 which is a successive count of the number of surface strips on each fruit f'or which the corresponding color rat-lo word exceeds a selectable threshold~ and a digital color size count signal on lines 3219 w~hich is a successive count of the number of` surface strips for each fruit.
The aforementioned five digital count signals are provided to the computer 409 ~l~ich analyzes the respective counts for each fruit to determine the grade catagor~J to whicil the fruit properl~J belongs. The com-puter 40 receives a sample signal on line 343 (shown ln FIG. 4) from the tim-lng un~.t B 171~ f'or tri~gerin~
the sampl.ing by the computer of~ the five count slgnals.
Each of the successlve pulses ln the sample slgnal occurs immediately af`ter a f'ruit has passed completely through the examining region 47~ and lmmedlately prior to a cor~esponding pulse in the reset signal on line 159~ which is used by the system to reset the respec~
tive counts to zero. At the time each sample pulse is 3 received, theny each of the five count signals will be a measurement derived af'ter the entire surf'ace of~ a ~ruit has been examined.
Pref'erablyg the c,omputer 40 normalizes the suc-cessive counts of the blemish count si~nal received on lines 275 b~J dividing them by the corresponding segmental counts of the segmental count or fruit size signal receLved on lines 3110 This results in a succession of~ blemish measures which represent the degree of the surface blemish on the f'ruit~ normalized : .~ .- . .
,~

',, ~ ~ .

r ~ i3 -42 ~
for size difrerencesO It will be apparent that ~hls normalization could JUst as readil~J be performed b~ a hard-wired digital dlvider circultO
Slmilarly; it is preferable to utillze the com-puter 40 to normalize the successive COUiltS of the colorcount slgnal received on lines 317 and the excess color count signal received on llnes 3199 by dividing them by the corresponding surface strip counts of the color count signal received on lines 321. T~s results in a 10 succession of measures of the average color of' the fruit and of the proportion of the surface area of each ~ruit having a color exceeding a predetermined selectable level.
The computer 40 is programmed with thresholds 15 defining the ~lemishg size and color limits of the various grade categories into which the frult are 'co be graded and sorted. The compu~er automatically compares the size count and the normalized blemishg color and excess color counts for each fruit to these thresholds 20 and determines the grade category in which the fruit properly belongs.
While the computer is performing the above-described operations, the ~ruit 21 are being transported by the second conveyor 2g from the examining region 47 25 to the sorting ~tation 28, Each o~ the solenoids 27 located at the sorting station 28 corresponds to a separ-ate grade categor~. I~en a solenoi~ 27 is actuatedg ~t tilts the portion of the conveyor 29 that ls immediately above it and thereby discharges any fruit thereon into 30 a receiver ~or a particular grade of fruit.
The computer 40 is also programmed with timing - lnformation which indicates the time that elapses while the fruit moves ~rom the examining region 47 to each of the solenoids 27 in the sorting station 28. At the 35 proper timesg the computer ou'cputs pulses on lines 345 to the appropriate solenoids, to discharge the fruit in accordance wlth the grade determlnatlons lt has made~
A computer is used to accomplish the above-described grading operations9 because i'c is ordinarily .

'' ~ 5~ ~ ~
-~13-readily reprogrammableg thereb~,7 permitting quick adapt,a-tion of' tl~e s~jstem tv accommodate di~ferences in fruit types and dif~erences in grading cakegoriesO Such dlf-~erences in gradillg cate~ories are ~enerallg due to changes in the markets to whlch the ~ruit are to be directedg and to variations in the fruit related to successive stages of the growing season.
It will be appreciated that t`ne specific program for the computer I~Q will depend on the selected fruit 10 grading crlteria ~or a particular situation. The compu_ ter may utillze the derived input parameters relating to blemish, size and color in any desired manner to sort and grade the ~ruitO An example Or a suitable computer program ~lowchart~ in simpl~ied ~orm~ is shown in FIG.
15 16.
~ rom the ~oregoing, it should be apparent that the present invention provides a new and improved method and apparatus for automatically grading and sorting fruit according to size5 surface blemish and surface 20 color. The apparatus utilizes a plurality of cameras for sequentially examining and generating re~lectance readings for a plurality o~ discrete areas on the ~ruit surface. The readings are suitably analyzed and com-bined to derive overall measurements of the slze~ blem-25 ish and color of the fru-~t. The fruit is then dis-charged to appropriate receivers in accordance wlth the measurements. The system is highl~J e~fectlve in pro-viding ~lexibllit~J to allow f'requent modi~ications to the grading categoriesO
While a specific rorm o~ the invention has been illustrated and described, ~t should be apparent that various modi~ications and variations can be made without departing from the spirit and scope o~ the inventionO
Accordingly, it is not intended that the invention be 35 limited, except as by the appended claimsO

'

Claims (16)

-44-

1. A method for measuring the surface color of an article, said method comprising the steps of: illum-inating the surface of the article with light having components within both a first and a second band of wavelengths; sensing light reflected from a plurality of unique areas on the surface of the article and producing for each of said areas a first measurement proportional to the intensity of reflected light within said first band of wavelengths and a second measurement proportional to the intensity of reflected light within said second band of wavelengths; comparing each of said first light intensity measurements to the corresponding one of said second light intensity measurements, thereby producing a plurality of characteristic color signals for the article; comparing each of the characteristic color signals to a predetermined threshold and producing color count pulses in accordance with the results of the comparisons; and counting the number of color count pulses, thereby producing a measure of the amount of surface having a prescribed color.

2. The method as claimed in Claim 1, further including the steps of: counting the number of unique areas on the surface of the article, thereby producing a measure of the size of the surface, and dividing the measure of the amount of surface having the prescribed color by the measure of surface size, thereby producing a measure of the proportion of the surface having the prescribed color.

3. The method as claimed in Claim 1, wherein said step of illuminating is performed when an article is disposed in an examining region, and wherein said method further includes the steps of: moving a plurality of articles in a sequential fashion through said examining region, whereby a separate measure of surface color is produced for each article;
and sorting the articles in accordance with the measures of surface color.

4. The method as claimed in Claim 1, further including the step of: averaging the plurality of char-acteristic color signals, to produce a measure of the average color of the surface of the article.

5. Apparatus for measuring the surface color of an article, said apparatus comprising: means defining an examining region; means for illuminating the surface with light having components within both a first and a second band of wavelengths; means for sensing light reflected from a plurality of unique areas on the surface of the article and producing for each of said areas a first measure-ment proportional to the intensity of reflected light within said first band of wavelengths, and a second measurement proportional to the intensity of reflected light within said second band of wavelengths; first means for comparing each of said first measurements with the corresponding one of said second measurements, thereby producing a plurality of characteristic color signals for the article; second means for comparing each of said characteristic color signals to a predetermined threshold and producing color count pulses in accordance with the results of the comparisons; and means for counting the number of color count pulses, thereby producing a measure of the amount of surface having a prescribed color.

6. The apparatus as claimed in Claim 5, further including: means for normalizing the measure of surface color in accordance with the size of the surface, thereby producing a measure of the proportion of the surface having the prescribed color.

7. The apparatus as claimed in Claim 6, where-in said normalizing means includes: means for counting the number of unique areas on the surface of the article, thereby producing a measure of the size of the surface;
and means for dividing the measure of the amount of surface having the prescribed color by the measure of surface size, to produce the measure of the proportion of the surface having the prescribed color.

8. The apparatus as claimed in Claim 5, further including: means for moving a plurality of articles in a sequential fashion through said examining region, whereby a separate measure of surface color is produced for each article.

9. The apparatus as claimed in Claim 8, further including: means for sorting the articles in accordance with the measures of surface color.

10. The apparatus as claimed in Claim 5, wherein:
said first comparing means includes means for dividing each of said first measurements by the corresponding one of said second measurements.

11. The apparatus as claimed in Claim 9, wherein:
said apparatus further includes means for averaging the plurality of characteristic color signals, to produce a measure of the average color of the surface of the article, and said sorting means is operable to sort the article in accordance with the measure of average color.

12. The apparatus as claimed in Claim 8, wherein:
said sensing means senses light received from said examining region and produces a plurality of pairs of said first and second measurements of light intensity for each article, each of said measurement pairs corresponding to a unique area on the surface of the article and including said first and second measurements.

13. The apparatus as claimed in Claim 12, wherein said sensing means includes: a plurality of phototransducer pairs arranged in a co-planar relationship with said examining region, each of said phototransducer pairs receiving light from a unique portion of said examining region and including a first phototransducer sensitive to light within said first band of wavelengths, and a second phototransducer, sensitive to light within said second band of wavelengths; and means for reading said phototransducer pairs in a sequential and repetitive fashion, thereby producing the plurality of pairs of light intensity measurements.

14. The apparatus as claimed in Claim 9, wherein said sorting means is operable to sort the articles in accor-dance with the measures of proportion of surface having the prescribed color.

15. The apparatus as claimed in Claim 7, further including: means for determining which of said pairs of light intensity measurements correspond to portions of each article, said determining means including means for comparing one measurement in each of said pairs of measurements to a pre-determined threshold, whereby if a measurement exceeds the threshold, it is determined that corresponding pair of measurements corresponds to a portion of an article.

16. The apparatus as claimed in Claim 13, wherein:
said examining region is a substantially planar examining region; said moving means includes conveyor means for moving the plurality of articles in a sequential fashion through said examining region; said plurality of phototransducer pairs are disposed on the periphery of said examining region;
said means for comparing includes means for computing the ratio between each of said first light intensity measurements by the corresponding one of said second light intensity measure-ments, thereby producing a plurality of color ratio signals for each article; means for comparing each of the signals in said plurality of color ratio signals for each article to a pre-established threshold, and producing said color count pulses in accordance with the outcomes of the comparisons;
and said sorting means includes means for sorting the articles in accordance with the measures of proportion of surface having the prescribed color.