JPS63163580A - Pattern analysis - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] An optical, digital, and electronic method for

The present invention particularly provides a symmetrical perceptive region.
The present invention relates to automatically determining the center of symmetry of the object by identifying the object (areas) and automatically determining its center of symmetry. It is also possible to arrange the recognition regions randomly within the scene and insert auxiliary patterns between them.

The method of the present invention that automatically identifies a recognition region and automatically searches for its center of symmetry is particularly applicable to an auxiliary pattern in which a pattern in a target scene is composed of symmetric recognition regions, and whose center of symmetry intersects the recognition region. is determined from the background of this scene by different gray levels, after observing the scene with a camera and digitizing the output of this camera, when the symmetric recognition area is geometrically different from the auxiliary pattern. different and whose auxiliary pattern, the transverse dimension of the recognition region is larger (at least locally) than the transverse dimension of the auxiliary pattern, and the symmetric recognition regions touch the recognition region by making them larger with respect to the auxiliary patterns that intersect them. It can be applied at any time.

BACKGROUND OF THE INVENTION Various methods are conventionally known as methods for automatically searching for symmetrical portions of a pattern. These methods are mostly used in the following situations.

Read plans, maps or drawings and use programs based on reference artwork to provide programming support to numerically controlled drilling machines, find reference points for automatic control robot 9Y and positioning systems.

In describing the present invention, the technical terms "digitization" and "pixel" will be used.

"Digitization" means representing the position and/or value of a pixel in digital form, especially for the purpose of electronic calculations.

The term "pixel" refers to a pixel that is conventionally expressed in the following format.

The smallest or basic unit of an image in digital form.

The basic unit in the basic structure of an image receiving sensor, such as a charge-coupled device (COD).

The basic unit of a digital image that is reconstructed on a screen or stored in memory after the image has been detected by a sensor and processed electronically.

The problem of automatically finding the center of a symmetrical shape has conventionally been solved by assuming that the recognition area of these symmetrical shapes is a simple shape (for example, a circle, square, etc.), and that these areas are distinguished from each other, and that the recognition area is a simple shape (for example, a circle, a square, etc.). This problem was resolved appropriately by preventing it from being combined with the auxiliary pattern that becomes .

For example, French Patent Application No. 8200706 (BOR
NE-LEC) describes a method and apparatus for reading artwork depicting pads of printed wiring circuits. This method can only be used for pads with a central point of symmetry, especially circular pads. Requires that there are no other types of auxiliary patterns in the scene. The method begins by determining the successive midpoints of line segments that intersect the edge of the pad in orthogonal X and Y directions. The method starts from a point located on the side of the pad. The horizontal extent of the pad is measured in a first direction (eg, X) at this level. A vertical bisecting line segment is then determined through the pad. The center of the pad is considered to be the midpoint of the above perpendicular bisector. This method could be accomplished using a single photocell moving in mutually orthogonal X and Y directions relative to the artwork. This method can thus be applied easily and effectively. However, this artwork is completely unusable in the normal case where it includes auxiliary patterns such as lines interconnecting pads.

Also, US Pat. No. 4,163,212 describes a method and apparatus for automatically identifying the centers of connection pads for integrated circuits. The recognition areas representing these pads are approximately rectangular in shape. The method described in this patent mainly consists of analyzing the scene in parallel lines along at least two orthogonal directions X and Y. For each of the directions X and Y, the curve formed by the center of the line segment formed by the analysis line intersecting this rectangular recognition area is determined. The center of this recognition area is
It is assumed to be the intersection of said curves representing centers along the various analysis directions. This method can be used to accurately determine the center of a recognition area using a simple pattern, such as a square area, provided that the recognition areas are isolated and independent within the scene. It is. If the scene contains other auxiliary patterns, this method will not work.

When it comes to recognition areas or auxiliary patterns for which the center of symmetry is to be determined, the prior art provides valid results only in the very rare case that these recognition areas contain identifying features.

In particular, French Patent Application No. 7,509,846 describes a method and a device for programmatically supporting numerically controlled drilling machines on the basis of artwork that includes printed wiring circuits. This method can be applied to include pads and interconnect lines. This method relies on giving a clearly distinguishable shape to the cognitive area to be identified. For example, this distinguishing shape can be made transparent by having a central hole in the middle of the pad to be drilled.

The image of the artwork is digitized within the window. The image is then analyzed and areas of the image that have a distinctive appearance are identified. In particular, the center hole of the pad is identified as well as the background area located between the interconnect lines (this is an error). Eventually, those areas that have a discernible appearance and touch the edges of the image window are extracted. In this way it is possible to identify only those recognition areas constituted by holes in the middle of the pad. This method works perfectly well a priori for artwork that has a pad with a central hole. However, this introduces errors and therefore cannot be used, for example, in artwork that has an etched area in the middle of the ground plane. This is because these areas become confused at the center point. Moreover, this method does not work at all in the more general case, such as a recognition area constituted by a pad without a central hole.

Similarly, U.S. Pat. No. 4,295,198 states:
A method and apparatus for automatically determining a center hole for a pad in printed wiring circuit artwork that includes interconnect lines is described. The method is based on a series of tests that successively confirm the agreement between various boundaries of increasing selectivity. The sequential sequence of these tests increases the speed of operation of the method. Each of these tests is performed in succession when a blank pattern is found. First test “BURST”
consists of checking whether the pattern is larger than the largest hole. This gives a very rough preliminaries.

The next test "L, PATH" is to move along the pattern according to the bus formed by the line segments perpendicular to the edges in order to check the maximum dimension condition. The test consists in erasing the line and thereby fulfilling the second selection. The final test “Pass (P
ATH)'' recognizes shapes of sufficiently small dimensions that have closed contours by tracing the contours.

It is assumed that this is a coming hole. This method is only applicable to detecting the pad by the hole.

Automatically determine symmetry recognition areas (e.g. pads) in the image of a scene (e.g. printed wiring circuit artwork) where the symmetry recognition areas are related to auxiliary patterns (e.g. interconnect lines), thereby determining the center of the recognition area. There is no general conventional method for determining this. As a result, there is still no known device that can automatically and reliably find a symmetrical recognition area, independent of the specific pattern, when the recognition area is related to an auxiliary pattern of arbitrary shape.

All currently available solutions either (a) require special patterns and thus considerably limit their usefulness;
(b) Either some type of identifier is required in addition to the pattern to be searched, resulting in the overall complexity of the device, or (c) The symmetrical recognition pattern is unusable if it happens to be combined with an auxiliary pattern. It had some drawbacks.

DISCLOSURE OF THE INVENTION The present invention seeks to solve the above-mentioned problems by automatically identifying randomly distributed symmetric recognition regions without risk of errors and automatically determining the coordinates of the center of symmetry of said recognition regions. , when attempting to find a recognition area in a scene that includes both the randomly distributed symmetric recognition areas and auxiliary patterns that meet and intersect these recognition areas,
The patterns constituting the recognition area (possibly combined with auxiliary patterns) are digitized (n for the recognition area, b for the background) to distinguish the scene from the background by the difference in gray level, ,
They consist of being at least locally distinguished from said auxiliary pattern by the fact that they have a transverse dimension larger than that of the auxiliary pattern, thereby constituting locally bulging shapes observed on said scene.

The present invention has the following advantages.

The exploration of symmetry areas is fully automated, such as automatic control of robots, automatic orientation of objects, plans, maps, etc.
It can be applied to many industrial applications such as automatic reading of drawings and automatic programming of numerical control machines.

The method of the present invention is universal and automatically applies to all kinds of scenes, even if there are arbitrary auxiliary patterns of various shapes associated with the symmetric recognition region, as long as the scene contains a symmetric recognition region. It can be applied to

The method of the invention is very fast and simple to operate, making it possible to use it in real time with a microprocessor. Furthermore, the method of the present invention allows execution time to be extremely short.

For this reason, the method of the present invention, after identifying a pattern in a scene by one gray level of its points, determines whether a center of symmetry exists by increasing the dimensions of the circles in the shape and by increasing the size of the circles within the shape. It consists of determining by determining the tangent circle. The general method of the invention is based on the following observation. That is, the largest circle that can be inscribed in a symmetrical shape always has its center on the center of symmetry of this shape.

DETAILED DESCRIPTION FIG. 1 shows a portion Z of an artwork 1 of a printed wiring circuit. In this artwork 1, the pattern on the transparent side is photosensitively made in opaque black, but the scene S constructed by artwork 1 is shown for example in a light box device (and the ground area F7 is all black in order to show it more clearly). , the background E of the scene appears white.

Printed wiring circuit artwork serves two important roles in the fabrication of printed wiring circuits.

First, the artwork l acts as a mask through which the photoresist on the copper epoxy substrate is exposed, leaving a pattern of pads, connecting lines and ground planes on the epoxy after development and etching, which A printed wiring circuit is constructed by this.

Second, artwork 1 is used as a measurement reference. That is, artwork l is used as a measurement reference when programming a numerically controlled machine that drills holes in a circuit to form holes for inserting component leads.

The method according to the invention can advantageously be used for dynamically programming the drilling of printed wiring circuits on the basis of such artwork.

A scene S composed of an artwork l includes a pattern f and an auxiliary pattern F constructed as follows.

That is, the pattern f is first, in this example, pad P1. . . . is constituted by a symmetrical recognizable region S to be identified, constituted by Ps, the central coordinates of which (for example the coordinates of C/1) must be determined for drilling.

Further, the auxiliary pattern F includes lines F+, . . . , F4 and the ground plane F7. It is composed of...

The pads P+, . . . , Ps are randomly arranged on the artwork l.

Auxiliary pattern F is not used for drilling programming, but simply pads P3. ..., to complete the process of searching for Ps. This is because these auxiliary patterns intersect the pads at connection locations 11.12.13.degree. 14, 15 and 16.

However, the pads P+, . . . have the width el+ e, +・of the line F,, F,, .
It is clear that the width d is locally larger than that of . As a result, the pads are bulged with respect to the auxiliary pattern that intersects them.

Also, pad P8. ... has a circular shape, so it is clear that it has the characteristic of being centrally symmetrical.

The automatic pattern analysis method according to the present invention is performed as follows.

First, a non-symmetrical auxiliary pattern F,...

From pad P1. . . . are automatically distinguished.

Next, the center C'1 . of the pad Pl, . . . ....

accurately and automatically.

To do this, in a first step a camera (not shown) is used to locate the part Z of the artwork that is to be analyzed.
Input an image window Ia covering . This portion Z includes pattern r. Figure 1 shows
One such window Ia is shown.

Then, in a second step, window 1a is digitized. That is, the electronic image of window Ia is stored in digitized form 1d. A portion of the image thus digitized is shown in FIG. 2'.

Different binary values are associated with the pattern f and the background E as a function of the gray level of the pattern r and the background E. In particular, the pattern f and especially the pad (Pl
) and the line (F, ), respectively (for example 20.21), are associated in this example with the binary digit zero, which corresponds to the color black.

Similarly, a square (for example, 23) belonging to background E is
It is associated with the number l, which corresponds to the color white. The electronic image processing method according to the invention is performed on a digitized image Id.

Then, in a third step, at least one first test point Co in each similar pattern f is fixed to form a portion of the symmetry recognition area P and in particular the pad P. A preferred method according to the invention for doing this is detailed below with reference to FIG. According to this method, the entire scene S
is scanned, and then all symmetrical recognition areas P in the pattern f described above are found.

This consists of multiple nodes N+, Nt with a constant pitch M.

N! , . . . , and the grid G is used to cover the entire scene S. Each node Nl, Nt, N3゜ of lattice G
... is the binary level (0
or l) is used as a starting point for a systematic analysis. The nodes N,, N,,... placed inside the pattern f and especially inside the pad P, are therefore
It is distinguished from the node N placed on the background. The node N, placed inside the digitized pattern fd, constitutes the next test point for performing the electronic pattern analysis method.

Therefore, the node NI placed inside the pad P is drawn centered on the point Co. Here, k=1゜2,...
・It is.

The concentric circles are increased in diameter dk. The concentric circles Cok are divided into digitized images 1d in different ways by juxtaposed pixels.
created within. For simplicity, these circles are drawn using dotted lines in FIG. In this way, circles Co' and Co' are drawn centering on test point Co.

Each day Co is divided into q arcs Qok1. Here, 1<q, and q corresponds to the number q of sectors Oo' distributed approximately around the center Co. In this case, q=
It is 4. The circle Co', Co'', --- therefore has four sectors or quarters qol+ l, QoI+ 2., centered around the test point Co.
, , Qo , Qo is divided into 1.

For each day CO, a test is performed to determine whether the circle Cok is completely contained inside the corresponding digitized pattern fd. As long as the circle Cok is included in the digitized pattern fd, here k k+1, in the case of the circle Co', the diameter d of the circle Co, Co,...
k is gradually increased.

On the other hand, one of the circles Co' in the group, say circle C
As soon as "o" intersects the digitized pattern "d, i.e. at least one point e of the circle co2 is outside the digitized pattern [d and therefore the digitized background E
As soon as it matches d, the digitized pattern fd
When only one arc qOLl of the circle Co'' intersects the digitized pattern (
In the seventh stage, the digitized pattern f
A new test point offset from the arc c, o* + 1 intersecting d is selected.

The procedure is then repeated from the new test point C1 through the fourth to seventh stages described above. A series of test points A C,, C,, C3,...C are created at this time. Often (as for test point C4) the first circle C413 has its two arcs q, +3+
1 and Q4'3□" occurs. Then, according to the invention, a new test point C6 is selected which is shifted from the two arcs of the problem intersecting the digitized pattern fd. The steps of the method described above are then repeated again.

From point C5, the sequence of test points Cn is its intermediate point C
You can see that it becomes fixed.

The use of the method according to the invention consists in considering that the pattern r can be analyzed to correspond nine times out of ten to the pad P, and assuming that the center of the navigation pad is C.

When trying to process a pad with two symmetrical axes, e.g. an oblong pad, symmetrically opposing arcs intersect the digitized pattern, while the sequence of centers Cn converges towards the center of symmetry. Something bad happens. To consider this situation (not shown), the number q of arcs is chosen to be an even number, for example 4. In each group of circles Cn, the arc Qnk1 is divided into symmetrically opposed pairs. The first circle Cn' of the group of circles Cnk centered on the test point Cn has a diameter dk
As we increase the digitized pattern ``d'', we consider symmetrical pairs of arcs, only one of which intersects the pattern. As long as at least one exists, a new test point C(.+1) shifted from the previous test point Cn by 1-) in the direction away from the arc intersecting such a pair of patterns is will be selected.

In any case, as shown in FIG. 3, as soon as the following conditions hold, a new set of test points Cn
is narrowed down within the digitized pattern fd.

Such conditions include, first, that there exists a circle Cnk that is centered on the aforementioned test point Cn and that is entirely contained within the digitized pattern fd; The next circle Cl1k+1 of the group is not less than the minimum permissible threshold value A (A≦q) and the number N of arcs Qnk+1,1 intersects the digitized pattern fd.
It is to have.

The test point Cn is assumed to be the hypothetical symmetry center C of the symmetry recognition area included in the pattern to be analyzed.

In FIG. 3, the number of arcs is 4 (q=4).

Additionally, the minimum acceptable threshold value A is also selected to be four. The center Cn satisfies the following two conditions. That is, the completely digitized pattern r is centered around the test point Cn.
The existence of a circle cnk located within d, the U circle Cn on O
k+' (D arc Qr+"", Qrlk+l"Q nk
+1. Child. nk+1.4, which intersects the digitized pattern rd.

The point Cn is considered to be the center of symmetry C of the symmetry recognition region P, which is assumed to exist at the intersection of the patterns f.

The performance of automatically determining the center of symmetry C using the method according to the invention is at the required level. This is essentially
The size of each pixel 20, 21..., the step movement size between successive test points Cn, C(1+1)'..., and the amount of change in diameter dk between successive test circles. Depends on it.

According to the invention, it is desirable to take the following considerations into account.

(i) The pixel size must be less than or equal to the precision S, (ii) The diameter increment d(k+1)-dk must be less than or equal to the precision S, (iii) Two consecutive test points Cn and C(n+
Particularly, referring to FIG. 2, the size of the basic Φ step during 1) is smaller than the precision S and the two test points C.

It is observed that the most common displacement between Qok1 and C1 is the vector v1, which can take any of four directions away from the arc Qok1 that intersects the pattern. In particular, the center Co
The amount of movement between and 01 is the digitized pattern f'd
The arc Qo that intersects with ! It is observed that the vector ■1 bisects the quota Oo' of l and the quota Oo'' on the opposite side.

If two non-opposed arcs of the first circle Cnk centered on test point 1 intersect the digitization pattern fd, the two elementary movement vectors are combined as shown for movements C4 to C2.

In most cases, the series of diameters dk chosen for the group of circles Cnk is the geometric sequence dk= (k+ ko)
Construct Do (same for each group of circles Cnk). is selected to be substantially smaller than the transverse dimension do of the smallest pad P. Declination angle Do is used when determining the center of symmetry (0 desired discontinuity (accuracy than S/JX) force)
, or equal.

Furthermore, the test grid G preferably has one edge e with the smallest pad P! It has a square mesh M smaller than the dimension do. In this case, the argument DO and the side e have the relationship e + Do < d.

It is advisable to satisfy , thereby ensuring that at least the first circle Co' of the group of circles centered at the test point CO lies completely inside each pattern to be analyzed.

A further confirmation test is performed when the hypothesized exact center of symmetry of the assumed recognition region P is found within the pattern f. The confirmation test will be described with reference to FIGS. 4, 5, and 6. This test is performed by analyzing the digitized pattern in at least one direction x-x', preferably two directions x-x' and Y-Y' using line-by-line scanning.

When scanned along the direction xx', the width of the digitized pattern rd is measured along the second direction Y-Y'. The change in width of the digitized pattern as a function of position g is shown in FIG.

Similarly, the change in pattern height 1-1 in the Y-Y' force direction is also investigated as a function of the xx' force direction (see Figure 5).

Since the pad P is larger than the line F intersecting this pad, each of the curves shown in FIGS. It has a second area 'lt, l'-.

According to the invention, it is advisable to confirm the assumed center of symmetry C simply by whether this requirement for increasing and then decreasing dimensions is confirmed in two different directions.

It is also advisable to perform a second maximum dimension test for an assumed center C. The pattern r including the pad P has maximum dimensions A and A' that are greater than or equal to the diameter dO of the minimum pad P. Therefore, the assumed center of symmetry is its dimension A,
A' is the minimum horizontal dimension do of pad P2 in both directions y, -X
It is checked whether the varying dimension (L, 1() of pattern r has an area H, H' that exceeds ', Y-Y').

Another way to check the assumed center of symmetry C is to find the diameter of the last circle Cn contained within the digitized pattern f'd by
Compare the horizontal dimension dO of the smallest pad P in the scene S, and similarly compare the dimension d(n+1) of the circle "(n+1) with the horizontal dimension D=d of the largest pad P. In this way, the hypothetical center of symmetry is
Confirmed only if the following relationships hold.

d > do and/or d(n+1) < D where the pattern f has a pad P with a center C and a diameter d. It can be considered that it includes the inscribed circle of .

Once the center of the annular pad P has been determined and confirmed by the above method, its diameter is then ascertained. For this purpose, the central dimension or size of the pattern, i.e. the spacing (
AI-CA' + ) in multiple directions, for example 4 directions (
at+ a' +), (at+ a' *), (a3+
a' j) and (a++ a' 4). The center size value (A, -CA' I) is within Y%,
At least N times the value A. (Y and N are predetermined parameters), the diameter of pad P is A. It can be considered that

It is used together with heart C to automatically detect it with the desired accuracy. This measurement is not influenced in any way by the presence of the auxiliary pattern F.

This method is particularly useful for use in a variety of fields requiring automated analysis of images and patterns. Additionally, it is particularly effective for automatically programming the drilling of printed circuits based on the artwork of the printed circuit.

[Brief Description of the Drawings] Figure 1 is a diagram showing a part of the artwork of a printed wiring circuit illustrating the principle of identifying shapes in a scene using the present invention, and Figure 2 is a diagram showing a part of the artwork of a printed wiring circuit according to the present invention. Fig. 2' is a diagram showing the details of Fig. 2 after it has been digitized, and Fig. 3 is a diagram showing the artwork of a printed wiring circuit that schematically shows how to find the center of symmetry. FIGS. 4, 5 and 6 illustrate the final stage of the method according to the invention when identifying, and FIGS. 4, 5 and 6 illustrate further properties of the method according to the invention for confirming the modification of the assumed center of symmetry. L...artwork, S...sheen, P...bara (and 1 other person) Fignia, head procedure amendment reading (method) April 28, 1985 Commissioner of the Patent Office Black 1) Akio Tono 1, case Indication Patent Application No. 300187 of 1988 2゜ Title of invention Pattern analysis method 3, HLI correction Relationship with each case Name of patent applicant Name Santol Robotic 10 agent Yoraga-cho, Chiyoda-ku, Tokyo 100- No. 519 Hibiya Park Building, 8-1 Chome (Telephone: 213-0686) 5 Date of amendment order March 31, 1985 6, Subject of amendment (+) "Brief explanation of drawings" column of the description and drawings Figure 2 (2) Corporate Certificate 7 Contents of amendment 1. The detailed statement will be amended as follows. (1) Page 34, line 19, “Figure 2” was corrected to “Figure 2A.” (2) On page 35, line 1, "Figure 2'" was corrected to "Figure 2B." ■Figure 2 of 3 drawings will be corrected as shown in the attached sheet.

Claims (1)

[Scope of Claims] 1. A system that electronically analyzes a pattern in a scene and automatically distinguishes symmetry recognition regions intersected by auxiliary patterns in the pattern and randomly arranged in the scene, The center of symmetry of the recognition area is determined together with the recognition area, and the auxiliary pattern is optionally digitized so that the background of the sticker can be distinguished by a gray level that generates different levels between pattern pixels and background pixels, Determine whether a pattern locally includes one of the symmetric recognition regions, use a camera to input a window of the scene containing the region in which the pattern is to be examined, and digitize the window into the pattern. digitized in the form of a digital action image containing the test points (Co
) is fixed, when a group of concentric circles with increasing diameters (dk) are drawn at their center positions with the test point as the center, and each circle Co^k is defined by q such that 1≦q, Divide it into q arcs Qo^k^l corresponding to q quarters 0o^l distributed around the test point, and as long as the group of circles is completely included in the digitized pattern, Continuously increase the diameter of the circle (Co^
k, Co^k^+^1,...), when a certain circle Co^i intersects the digitized pattern (fd), that is, circle C
At least one point (e) of o＾i is the pattern (fd)
As soon as the circle Co^i lies outside of and coincides with the digitized background, it intersects the digitized pattern (fd).
If only one arc of the circle intersects the digitized pattern, the number q0 of arcs Qo^i^j of is determined, and if only one arc of the circle intersects the digitized pattern, the direction opposite to the direction of said arc Qo^i^j that intersects the pattern is determined. The new test point C_l moved from the previous test point to
by selecting a test point and repeating this step from that new test point as long as each circle Cn^k that intersects the digitized pattern continues to such intersect with only one of its arcs Qn^k^l. If the central sequence converges at a limited number of points and becomes fixed near such a point, then the pattern constitutes a symmetric recognition region and is attached to said point. A pattern analysis method characterized by assuming that there will be a center. 2 The number q of selected arcs Qn^k^l is an even number, especially the circle C
Equal to 4 in all groups of n^k, the arc Qn^k^
Claim 1, characterized in that l is defined by a quarter 0n^l in a certain direction, and the arc Qn^k^l of each circle Cn^k is divided into symmetrically opposing pairs. the method of. 3. Whenever the first circle Cn^l of the group of circles Cn^k centered at the test point (Cn) intersects the digitized pattern (fd) during the analysis with increasing diameter, the symmetrical pair of arcs Determined if only one arc of the symmetric pair intersects the digitized pattern, and if at least one pair of arcs of this symmetric intersection type exists, one intersecting arc of said symmetric pair of arcs. 3. A method as claimed in claim 2, characterized in that a new test point is selected that has been moved from the previous test point in a direction opposite to the direction of . 4. A series of test points Cn in a digitized pattern (fd) is centered on said test point Cn such that the series of test points Cn satisfies the following conditions: first, it lies entirely within the digitized pattern being analyzed; Place the circle Cn^
k, Second, the next circle C in the group centered at the test point Cn
The number N of arcs Qn^k^+^1^,^l where n^k^+^1 intersects the digitized pattern (fd) larger than the minimum allowable threshold value A, when A is A≦q. The test is stopped when the following conditions are satisfied, and the last test point Cn is the center of symmetry of the symmetry recognition area of pattern (f). Method described. 5. A series of test points Cn in a digitized pattern (fd) is centered on said test point Cn such that the series of test points Cn satisfies the following conditions: first, it lies entirely within the digitized pattern being analyzed; Place the circle Cn^
k, Second, the next circle C in the group centered at the test point Cn
All arcs of n^k^+^1 Qn^k^+^1^, ^1
intersects the digitized pattern (fd),
5. The method according to claim 4, wherein the test is stopped when the condition (f) is satisfied, and the last test point Cn is set as the symmetry center of the symmetry recognition area of the pattern (f). 6 The series of diameters (dk) of groups of circles Cn^k are the same geometric sequence dk=(k+k0)D0 for each group of circles.
and the constant k0 is the first diameter d1=(1+k0)D
The transverse dimension d of the minimum symmetric recognition area that 0 is trying to find
2. Method according to claim 1, characterized in that the argument D0 of the geometric sequence is selected to be smaller than 0 and is substantially equal to the absolute precision s required for automatic determination of the center of symmetry. 7 Requiring precision s, the elementary movement between two consecutive test points constitutes a vector (C_n, C_(_n_+_1_)), which intersects the digitized pattern with an arc Qn^k^+^1 Opposing ^, ^l and arc Q
Claim 1, characterized in that it is substantially oriented in the direction n^k^+^1^,^l and has a constant length l equal to or smaller than the required absolute precision s. The method described in section. 8 The digitized image of the entire scene is scanned using a constant mesh to ensure that the entire scene has been scanned and that all symmetry-recognized regions included in the pattern of that scene have been detected along with their respective centers of symmetry. is systematically analyzed at a plurality of test points located at the nodes of a test grid to determine whether the binary levels represent background or pattern gray levels, and said grid is digitized. 2. A method as claimed in claim 1, characterized in that over the entire scene, each of said nodes is successively taken as a first test point if its binary level matches a portion of the pattern. 9. Method according to claim 8, characterized in that the mesh of the test grid is a square with sides e<d0, d0 being the smallest transverse dimension of the symmetry recognition region. 10 To avoid errors due to the presence of auxiliary features intersecting the symmetry recognition area, the digitized image in at least one direction Observe the change in the width of the digitized image of ', and confirm the existence of the assumed symmetry center and the assumed symmetry recognition area only if the width of the pattern in the Y-Y' direction includes steps of increasing and decreasing. A method according to claim 4, characterized in: 11. In order to check the assumed center of symmetry of the symmetry recognition area, the assumed symmetry center is checked only if the digitized pattern contains a part with a width greater than the transverse dimension d0 of the minimum symmetry recognition area of the scene, and that part 11. A method as claimed in claim 10, characterized in that is between an increasing phase and a decreasing phase. 12 For each assumed center of symmetry, compare the diameter dn of the final circle Cn contained therein with the minimum transverse dimension d0 of the minimum symmetry recognition area of the scene, and/or calculate the diameter of the next circle C_(_n_+_1_) d_(_n_+_1_)
is compared with the transverse dimension D0 of the maximum recognition area of the scene, and the assumed center of symmetry is dn≧d0 and/or d_(_n
＿＋＿１＿）≦D0 only if the center of symmetry is confirmed to be accurate, or the pattern (f) is the determined center (C)
5. A method as claimed in claim 4, characterized in that a similar comparison is made in the case of a symmetrical recognition region having the same diameter as the inscribed circle. 13 After the center of the recognition area has been determined and the determination of the size of the center of the digitized pattern in R directions has been confirmed, the distribution of the size of said center is substantially analyzed and the diameter d of the symmetric recognition area P is determined. is the value of the center size is at least N times the value A
2. A method according to claim 1, characterized in that A is considered to be equal to A if it is considered to have (up to y%).