JPH0792441B2 - Pass / fail judgment method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pass / fail judgment method in an inspection in which it is difficult to determine a pass / fail threshold such as an appearance inspection using a grayscale image.

[0002]

2. Description of the Related Art In a conventional main inspection such as a defect inspection of a printed circuit board or a displacement inspection of a component, a characteristic difference between a non-defective product and a defective product is remarkable, and a pass / fail judgment is made by a binary image feature amount of an inspection unit. It was possible. Therefore, in the conventional quality determination method, the feature amount threshold value (not the threshold value for binarizing an image) for quality determination is fixed. For example, as a method for inspecting a fixed threshold value, Japanese Patent Laid-Open No. 62-17768.
There is publication No. 1.

[0003]

However, in the soldering inspection of the surface mount pin grid array and the inspection of the narrow portion by the fiberscope or the like, it is necessary to process the grayscale image and the poor image and judge the quality. Therefore, it is impossible to perform the optimum quality determination with the fixed threshold value as in the conventional quality determination method.

Further, in the inspection by an inspection object having a slightly different shape such as soldering, or a fiberscope in which the relative distance and posture with respect to the inspection object cannot be precisely positioned, the quality judgment based on the fixed threshold value is not possible. It was possible.

An object of the present invention is to provide a method capable of making an accurate pass / fail judgment according to the variation in the distribution of the data group of the feature quantity to be inspected. In particular, the present invention provides a method capable of making a stable pass / fail determination even when there is a large variation in distribution, which cannot be accurately determined by the conventional method.

[0006]

SUMMARY OF THE INVENTION The present invention is a pass / fail judgment method in an inspection in which it is difficult to determine a pass / fail threshold such as a visual inspection using a grayscale image. Feature quantity input means for taking in the feature quantity extracted from the image or the like as array data, and a feature quantity change calculating means for calculating the change quantity relating to the sample number of the feature quantity taken in by the feature quantity input means. A feature amount change histogram generating means for generating a feature amount change histogram calculated by the feature amount change calculating means, and a histogram for calculating a half width of the histogram generated by the feature amount change histogram generating means A half-value width calculation means, and a quality determination means for performing quality determination based on the half-value width calculated by the histogram threshold price width calculation means A quality determination method, characterized in that it comprises.

[0007]

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a quality determination processing system of the present invention. As shown in FIG. 1, the quality determination processing method includes a feature quantity input means 11 for inspecting N inspection objects and taking in the feature quantity extracted from an image of an inspection unit or the like as array data. A feature amount change calculation unit 12 for calculating a change amount of the feature amount fetched by the input unit 11 and a feature amount for generating a histogram of the feature amount change calculated by the feature amount change calculation unit 12. Change histogram generation means 13
And a histogram half-value width calculating unit 14 for calculating the half-value width of the histogram generated by the feature amount change histogram generating unit 13, and a pass / fail judgment based on the half-value width calculated by the histogram half-value width calculating unit 14. And a pass / fail judgment means 15.

The feature amount input means 11 inspects N inspection objects, and the feature amount f extracted from the image of the inspection unit or the like.
(I) is taken in as array data. However, i is a sample number and i = 1, 2, ..., N. For example, 30
When individual inspection objects are inspected, the characteristic amount f (i) as shown in FIG. 2 is obtained. Here, the inspection object of sample number 7 is a defective product.

The characteristic amount change calculating means 12 is characterized by the characteristic amount f
(I) is processed by the following calculation and
Calculating feature quantity variation df a (i). When i = 1 df (i) = f (i)-{f (i + 1) + f (i +
2)} / 2 When i = 2 df (i) = f (i)-{f (i-1) + f (i + 1)
+ F (i + 2)} / 3 When 3≤i≤N-2 df (i) = f (i)-{f (i-2) + f (i-1)
+ F (i + 1) + f (i + 2)} / 4 When i = N-1 df (i) = f (i)-{f (i-2) + f (i-1)
+ F (i + 1)} / 3 When i = N df (i) = f (i)-{f (i-2) + f (i-
1) } / 2 df calculated according to the above from f (i) shown in FIG.
(I) is shown in FIG. The feature amount change rate df (i) is distributed with 0 as the median value.

The feature amount change histogram generating means 13 is
A histogram having the characteristic amount change df (i) calculated by the characteristic amount change calculating means 12 on the horizontal axis and the frequency on the vertical axis is generated. FIG. 4 shows a histogram generated based on the characteristic amount change df (i) shown in FIG.

The histogram half-value width calculating means 14 uses the least squares method to form a Gaussian function having coefficients a1 and a2: F (x) = a1exp (-41n2x2 / a22) = a1exp (-2 .773 · x2 / a22) to the histogram generated by the feature amount change histogram generation means 13. The value of the parameter a2 obtained as a result is the half-value width F of the combined Gaussian function.
Then, the magnitude represents the magnitude of variation in the feature amount change rate df (i). The result of fitting the Gaussian function to the histogram shown in FIG. 4 is shown in FIG.

The quality determination means 15 determines quality by the quality determination method described below using the half width Γ calculated by the histogram half width calculation means 14.

If | df (i) | ≦ aΓ, the i-th sample is a non-defective product. | Df (i) |> aΓ, the i-th sample is a defective product. However, || is an absolute value and α is a positive value. It is a real value. α is
This is a parameter that represents the severity of the non-defective item determination. The larger the value, the weaker the determination. In FIG. 6, α =
A graph of the characteristic amount change df (i) in which the pass / fail judgment threshold value Γ when 1 is shown by a dotted line is shown. It can be seen that the 7th defective product is correctly determined.

[0015]

According to the present invention, the threshold value for the quality judgment is explained according to the variation of the distribution of the data population of the feature quantity to be inspected. Even if the variation in the distribution is large, it is possible to accurately determine the quality.

As described above, according to the present invention, it is possible to stably carry out the quality judgment in an inspection in which it is difficult to determine a quality judgment threshold value such as an appearance inspection using a grayscale image.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of the present invention.

FIG. 2 is a feature amount f when 30 inspection objects are inspected.
It is a figure which shows the example of (i).

FIG. 3 is a diagram showing a characteristic amount change df (i) calculated by processing the characteristic amount f (i) shown in FIG. 2 by a characteristic amount change calculating means.

FIG. 4 is a diagram illustrating a feature amount change histogram generation unit of FIG.
It is a figure which shows the feature-value change histogram produced | generated by processing the feature-value change df (i) shown in FIG.

5 is a diagram showing a result of fitting a Gaussian function to the feature amount change histogram shown in FIG. 4 by a histogram half-value width calculating means.

FIG. 6 is a characteristic amount change df indicated by a dotted line with a threshold value when α = 1 calculated by a histogram half-value width calculation means.
It is a figure which shows (i).

[Explanation of symbols]

 11 Feature Amount Input Means 12 Feature Amount Change Calculation Means 13 Feature Amount Change Histogram Generation Means 14 Histogram Half Value Width Calculation Means 15 Good / Failure Judgment Means

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location G06T 7/00

Claims (1)

[Claims] 1. A pass / fail judgment method in an inspection in which it is difficult to determine a pass / fail threshold such as a visual inspection using a grayscale image, wherein N inspection objects are inspected and extracted from an image of an inspection portion or the like. A feature quantity input means for loading the quantity as array data, the feature quantity input means, and a feature quantity change calculation means for calculating a variation quantity related to the sample number of the feature quantity fetched by the feature quantity input means, A feature amount change histogram generation unit for generating a histogram of the feature amount change calculated by the feature amount change calculation unit, and a histogram half for calculating the half width of the histogram generated by the feature amount change histogram generation unit. A price range calculating means and a quality determining means for performing quality determination based on the half width calculated by the histogram half width calculating means. A pass / fail judgment method that is characterized by