JPH03218425A - Method and device for discriminating color difference - Google Patents

Abstract

PURPOSE:To improve reliability in discriminating a color difference by conducting fuzzy color difference discrimination when the color difference value of a body to be measured exists between two preset values and performing binary color difference discrimination in other cases. CONSTITUTION:The measured color difference value of a body from a color camera or a color sensor is fetched in CPU through an interface, the input voltage for the color difference is transduced into a color space indicated value quantitatively expessing colors colors which is compared with the preset value. Namely, when the color difference value of the body to be measured exists between the two optionally preset values, fuzzy color difference discrimination is conducted, and OK or NO binary color difference discrimination is performed when the value exists beyond the two set values. Accordingly, the subtle color difference such as the color difference of a woven fabric can be discriminated in the same way as that conducted by a human, and the reliability in discriminating the color difference is enhanced.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Use) The present invention relates to a method and apparatus for determining the color (color difference) of an object using an industrial color camera or color sensor.

(Prior Art) Color discrimination of objects has already been carried out in various ways by using industrial color cameras and color sensors. For example, measure the color difference (color difference value) of an object, compare it with a preset value, and if the color difference value (ΔE) is larger than the set value, N
A binary color difference determination is performed in which the value is OK if O is smaller.

(Problem to be solved by the invention) Color difference determination is always OK due to ambiguity in the color of objects
．． In many cases, accurate and reliable color difference discrimination cannot be performed using a binary judgment of NO. In particular, when inspecting and distinguishing uneven dyeing of fabrics or subtle color differences between fabric lofts, the current practice is to perform visual inspections by humans to improve reliability.

However, there are variations in human judgment over time.
It is often insufficient in terms of reliability and uniformity.

(Means for Solving the Problems) The present invention solves the problems of the conventional technology, and provides a method with excellent uniformity and reliability that can discriminate within an accepted range for determining subtle color differences. and equipment.

That is, the present invention provides a color difference determination method for comparing and determining the color of an object with a reference color, in which a color difference value of an object to be determined is measured, and the color difference value is set to an arbitrarily set value ε1 in advance.
When the color difference exists between ε1 and ε2, a fuzzy color difference determination is performed, and when the color difference exists other than that, that is, between ε1 and 62, a binary color difference determination of OK or NO is performed. This is a determination method and a color difference determination device for comparing and determining the color of an object with a reference color. ε2, and performs fuzzy color difference judgment when the measured color difference value exists between ε1 and ε2, and performs binary judgment of OK or NO when the measured color difference value exists in the other range. This is an object color difference determination device characterized by the following.

The present invention uses color differences (color difference values) to discriminate the colors of objects, and is preferably applied to online color discrimination, and is based on the uniformity of equipment (no sagging over time) and the comprehensiveness of humans. The present invention provides a method and apparatus for color discrimination of objects, which is both accurate and has excellent points that can be determined within an acceptable range of ambiguity.

In other words, the range in which OK and NO can be clearly determined, and the range in which OK and NO can be clearly determined.
For the latter, set values are set according to the color and material of the object to be measured, and when there is a measurement color difference between the set values, fuzzy theory is used. This is what we are trying to determine using.

Hereinafter, explanation will be given using the C I E ; L, a, b color system as an example. The measured color difference value of an object from a color camera or color sensor is converted into a quantitative value by a calculator (
computer).

On the other hand, using a fuzzy set that can numerically express the judgment criteria of humans who make subtle and ambiguous comprehensive color difference judgments, membership functions are created from the data necessary for color difference judgment, and the color difference of objects is calculated online. It is used for judgments in boundary areas where the ambiguity when making judgments is particularly significant. Therefore, the measured color difference values (ΔE) have set values ε1 and ε
A fuzzy color difference determination is performed only when the value is between ε1 and ε2, and a binary color difference determination of OK or NO is performed when the value is other than between ε1 and ε2.

An example of the system conceptual diagram of the device according to the present invention is shown in the first diagram.
Shown in Figure 3.

(Examples) Example 1 FIG. 1 shows the system configuration of the online color difference discrimination device of the present invention. Color difference values (color values) are data input from a color camera or color sensor and are taken into the CPU through an interface. At this point, the input voltage value of the color difference value is converted into a color space display value that quantitatively represents the color. Here, the description will be made using the CIE Lab color system. The CPU also controls the online drive unit and turns on/off the power of the conveyance system for the colorimetric object.

Additionally, this system includes a notification device (e.g. buzzer, abnormal lighting lamp, etc.) and a recording device (
Figure 2 shows the operation flowchart of this case, which consists of peripheral devices such as a hard disk, floppy disk device, etc.), a printer, a display, and a keyboard. Online color difference determination will be described in detail below.

When the system is powered on and becomes operational, initial value settings are performed. (Here, initial value setting refers to setting the measurement of the color difference meter as L, a, b color system, inputting the data value Lo + ao + bo of the reference color, etc.) Here, we will explain the basics of color difference discrimination. Only the necessary parts will be explained. In other words, in this example, the mode is for measuring color differences, and in order to make online color judgments, reference color data is used alone.

It is assumed that this reference color data is set to a value (Lo+ao+bo) by inputting it from the keyboard or embellishing the reference object.

At this time, the setting of the color difference determination value differs depending on the input value. For example, the degree to which humans perceive color differences differs depending on red-based colors or blue-based colors, so the color difference value (ΔE) is
b This shows that the color is not uniform throughout the color space.

In other words, it is necessary to change the color difference determination value ε7 according to the input value of the reference color data. The setting of this color difference judgment value changes depending on the object whose color is to be measured, the purpose of the measurement, etc., and may also change depending on the person making the final judgment. Therefore, by applying the fuzzy set concept to this part, we created a membership function based on the inspector's know-how and experience, and used this to make the same judgments as conventional human inspections. A reference value for color difference determination is calculated using fuzzy inference. (Membership functions and inference methods will be described later) The color difference determination values (ε, .) are the color values L, a, b of the object to be determined and the color value L of the reference color. +aO+l) This value is used to determine whether or not there is a color difference by comparing it with the color difference value △E from When NO, ΔE≦ε, there is no color difference and it is OK.

In addition, the reference value for color difference judgment is, for example, for red, ΔE=2
．． 0, there is a color difference with a value close to ΔE = 0.7 for dark blue,
There is a value based on conventional human judgment that says ``No,'' but the acceptable value for color difference judgment differs depending on the color and color system.Here, the standard color L o r a o +
It is necessary to set a reference value for color difference determination in advance for the hue of bo, and the value is different for determination depending on different reference colors.

Since it is derived from L o + a o , bo, the reference value for color difference determination is ε, here. It is expressed as

Therefore, in the present invention, ε7 is ε1. Setting values ε and ε2 are set on the tenth side and the first side with , as the center.
If ΔE is smaller than ε1, there is no color difference. If ΔE is larger than ε2, it is determined that there is a color difference. If ΔE is between ε and ε2, the fuzzy theory is used for determination.

Create a rule to obtain a reference value for color difference judgment from the membership function. That's it. As an example, IF L is dark (AND) a is red (AND) b is yellowish THEN color difference judgment reference value ε. We will create a rule that takes a value that seems to be 1.0. In creating this membership function and rules, we bring in the know-how and experience of inspectors who actually perform color inspections.

Next, also in the case of color difference determination fuzzy as the second step, membership functions are configured as shown in FIG. Here, in order to perform color difference judgment, the input values are (△L, Δa,
4(a), (b). A membership function like (c) can be considered. Further, the final color difference determination also shows three types of membership functions: it seems that there is no color difference, there is some color difference but it is almost OK, and there is a color difference.

The rule creation is similar to the previous section, for example, IF ΔL is slightly brighter < (AND) Δa is approximately the same Δb is approximately the same THEN Although there are some color differences, it is almost OK.

A rule like this is created.

Specifically, in Figure 4, △L: 3 ways, Δa: 3 ways, △
Since there are 23 ways, the number of cases is 3X3X.
3 = 27 ways, and 27 rules can be created.

FIG. 5 shows an inference method for fuzzy color difference determination.

This method is a direct method that is often used in fuzzy control. To explain this diagram, first, when input values ΔL, Δa, and Δb for determination are obtained, ΔL1Δ
It is determined which rule the value of a1Δb corresponds to among the 27 rules mentioned above. Then, the rules necessary for fuzzy inference calculations are selected. Let's now call this the rule ■, ■. Then, the consequent proposition is truncated from the grade values of ΔL, Δa, and Δb, and a membership function is obtained by combining them. Then, the center of gravity value is 1. is calculated.

This ε. is the fuzzy set that belongs to the highest grade of the membership function, that is, in this example, it is determined that "there appears to be a color difference." Through the above procedure, the presence or absence of color difference is determined by fuzzy inference. (For fuzzy inference methods, see "Fuzzy Control" by Michio Kanno (Nikkan Kogyo Shimbun), pages 76-83.
detailed in. ) The system with the above configuration was experimentally verified, and it was confirmed that the system has superior determination efficiency compared to the binary color difference determination method. This will be explained below.

The color sensor used in the experiment was a Minolta color difference meter CR-210, which was connected to a personal computer to perform operations such as data input, processing, and discrimination.

In addition, the colorimetric object we selected was a dark blue color that is used for fabrics that require delicate color difference determination, especially for men's suit fabrics. At the same time, we collected data on color difference limit samples that were actually judged, and created a membership function for color judgment based on this data. FIG. 6 shows limit color difference sample data of a navy blue system, and (a) is shown as an a-b plane, and (b) is shown as an a-L plane view. Figure 7 shows the membership function created from Figure 6, and rules were created based on this. On the other hand, the membership function for color discrimination is
As shown in the figure. The base set (variable definition range) of each membership function of ΔL, Δa, and Δb was set using color difference determination reference values ε and 0. The above fuzzy inference color determination operation was implemented as software and an experiment was conducted.

First, the method that calculates the reference value for color difference determination using fuzzy inference obtained results that were almost 100% as expected.

This was investigated using critical color difference sample data as input values.

Next, the Fuzzy inference value for color difference determination also obtained almost the same results as those determined by humans.

FIG. 9 is a diagram illustrating the reliability of the binary color difference determination method and the fuzzy color difference determination method. (a) is binary color difference judgment;
In (b), the horizontal axis is the color difference value when performing the vuzzy color difference determination. (a) is the guideline value ε,. If it is smaller than o, it is OK; if it is larger, it is NO. If we consider the measurement variation in color sampling and consider it probabilistically, the measured value ε,. If we take the neighboring values, we are making an OK/NO decision with a reliability of 0.5 (50%). On the other hand, in fuzzy color judgment, the final question is whether there is a color difference or not, so the fuzzy set is also ``There is some color difference, but it is almost OK.''
``There are 4 color differences'' and 2.
It will be determined based on one set.

If the value obtained by fuzzy inference is a point on line segment BC in figure (b), it cannot be determined that it belongs to either set when it exists within △PBC (indicated by diagonal lines). . In other words, if it belongs to △APB or △DPC, it can be determined whether there is a color difference or not, so when a point on BC is given, the reliability of the judgment is: Area ratio = (△APB + △DPC) / APDCB = Accurately Area where color difference can be determined/Points on BC exist and are given in the entire vine grade area.

In this experiment, all membership functions are expressed as straight lines (for example, AC.fir), and the intersection point P has a grade value of 0.5, so the reliability is 0.88.
7 (88.7%). This is only the final 2
Although the calculation is based on the assumption that color difference determination is performed, fuzzy color determination is superior in terms of reliability. If a membership function is created that reduces this ΔPBC, the reliability will be further improved.

As described above, color difference determination could theoretically be realized with higher reliability than the binary color difference determination method, even though the reliability may be affected by the shape of the membership function.

Example 2 A different model from Example 1 was used to create rules for fuzzy inference.

If we take a closer look at how humans perform color difference judgment, for example,
They make judgments such as "There is a color difference because it has a slight reddish tinge" and "There is a color difference because it is a little bright and yellowish." Considering this in the color space of Lx a book bx system, each Lx ax
A standard is set for each axis, and the determination is made by comparison with that standard. In other words, for the reference color (L1, a1, b o"), Lo"±εL on the LX axis and a1±εa1 on the a'' axis.
b on the bx axis. This corresponds to determining reference values ε, , εa1εb for color difference determination, such as X±εb. Then, if the value of the measured point falls within the area on the color space defined by ε, εa1εb (this can be considered as an ellipsoid), it is considered that there is no color difference, and if it is outside that area, it is considered that there is a color difference. (Figure 10).

In this case, the rules for finding the reference values for color difference judgment can be created in the same way as in Example 1, but
Since it is necessary to obtain the value for the bx axis, there are three types of rules.

That is, {IFL” value is Li r a” value isaLblI value is bi; THEN gL isgi} {IF
L” value is Li + a” Mis al b+
1 value isbi ;T}! EN εais gai)
{IF L" value isLl.a' value 1sai.b" value i
sb[;T}IEN gbis gbi} (N.
M. P: integer) Therefore, the fuzzy inference in the first stage is based on the reference color (L
ot, ao” + box), the reference value for color difference judgment at that point is εLa in the LX axis direction and ε in the grass axis direction.
It is shown that in the a+b direction, the value is εb. Next, a method of determining color difference when the color value (Lkyo,a''+bx) of the measured point is obtained will be explained.

The rules in this case are different from the first embodiment and are configured as follows.

N (IFΔLisΔLl, ΔElsΔEI THEN
f Isfl) {IFΔaIsALi +ΔatS
ΔEi THEN f 1sf1}P {IFΔb1sΔlj, Δb1sΔEI THEN
f isfl}+si where ΔL=LX-Log, Δa=a”
1a O”+Δ b=b book −bog Δ Er
(ΔL)”+(Δa)”+(Δb)2. Further, f is an index used when determining whether there is a color difference. This rule is established so that ΔL, Δa, and Δb can be used to judge the color difference independently, but when the measured point is at Q as shown in Figure 11, Δa, εa, Δb, and ε
b, but it is considered appropriate to determine that there is a color difference because it is outside the area of the elliptical sphere that serves as a standard for color difference determination. Therefore, since the distance δ from the point P is the boundary of the elliptical sphere, the color difference can be determined by examining the relationship (size) between PQ=ΔE and δ.

However, since δ is determined by ε, εa, εb, ΔL, Δa, and Δb, rules cannot be created using only ΔE and δ alone.

From the above explanation, the fuzzy rule for color difference determination has the above-mentioned configuration.

Here, as shown in FIG. 12, it is expressed as a numerical value of [0.1]. Therefore, it was confirmed that the configuration of fuzzy color difference determination using this ellipsoidal sphere model also matched visual determination with an extremely high matching rate, as in Example 1.

(Effect of the invention) When determining the color difference of an object, especially when determining the color difference of an object online, because the determination method is ambiguous, it is difficult to judge subtle color differences such as the color difference of textiles using the same method as humans. The reliability of the judgment is higher than that of the conventional binary color difference judgment method of OK/NO. Therefore, the efficiency of online color difference determination can be improved.

[Brief explanation of drawings]

FIG. 1 is a schematic diagram of a system configuration in an embodiment of the present invention, FIG. 2 is a schematic diagram of an operation flowchart of the embodiment, and FIG. 3 is a schematic diagram of an example of a membership function in the first embodiment. , FIG. 4 is also a schematic diagram of the membership function example in Example 1, FIG. 5 is a schematic diagram of the inference method for fuzzy color difference determination in Example 1, and FIG. 7 is an example of the membership function in Example 1, FIG. 8 is an example of the membership function in Example 1, and FIG. 9 is the conventional binary judgment method and the present invention. 10 and 11 are schematic diagrams for explaining the reliability with respect to the method, and FIGS. 10 and 11 are ellipse models in color space. FIG. 12 is a schematic diagram of an example of a membership function in the second embodiment, and FIG. 13 is an example of a conceptual system diagram of an apparatus according to the present invention.

Claims (2)

[Claims] (1) In a color difference determination method for comparing and determining the color of an object with a reference color, when the color difference value of the object to be determined is measured and the color difference value exists between preset values ε_1 and ε_2, fuzzy A method for determining color difference of an object, characterized by performing color difference determination, and performing binary color difference determination of OK or NO at other times. (2) A color difference determination device for comparing and determining the color of an object with a reference color, an element for measuring the color difference value of the object to be determined, setting values ε_1 and ε_2 for determining the color difference value, If the color difference value exists between ε_1 and ε_2, fuzzy color difference judgment is performed, otherwise OK or N
A color difference determination device comprising at least an element for performing binary color difference determination of O.