JP2013032995A - Checking program, recording medium having the same stored therein and checking device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a checking program which has no need of selecting a nondefective beforehand in order to objectively set tolerance as a statistically meaningful value, and is independent of the ability of an operator.SOLUTION: A computer is made to execute: a first statistic step of calculating a statistic of characteristics for each pixel having the same coordinates and preparing a provisional good pixel range for a digital image group individually acquired from the prescribed number of a plurality of objects of the same specification; a storage step of storing the digital image group in a memory; a temporary determination step of determining whether or not the characteristics of each pixel belong to the provisional good pixel range for each image in the stored digital image group; a second statistic step of calculating the statistic of the characteristics for each pixel determined as belonging to the provisional good pixel range, and preparing a set good pixel range; and a main determination step of determining whether or not the characteristics for each pixel belong to the set good pixel range for a digital image acquired from the object.

Description

  The present invention relates to a program for inspecting the presence or absence of defects such as scratches and defects on an object from an image obtained by imaging the object, a recording medium storing the program, and an inspection apparatus.

  When the quality of an object is determined using a digital image obtained by imaging the inspection object, an image acquired from an object that is known as a good product in advance is compared with an image acquired from the inspection object. There is a way. In this method, it is necessary to set a large tolerance or to set a different tolerance for each pixel in order to allow variation of the non-defective product itself. Also, a method has been proposed in which importance is set for each pixel so that a difference from a reference image in a less important part is not recognized as a defect (Patent Document 1).

  In any case, since tolerance and importance are set based on the subjectivity of the worker, the worker is required to have advanced knowledge and experience that can understand all defects, and the number of workers that can be set is limited. In addition, when optical characteristics such as luminance for each pixel are used as the inspection medium, the reliability of the pass / fail judgment is reduced with the illuminance fluctuation of the illuminator.

  Therefore, the inventor calculates the average and standard deviation of the brightness for each pixel having the same coordinates for the digital image group acquired from a plurality of non-defective products selected in advance, and creates an evaluation formula from these, A method for determining pass / fail based on the luminance and the evaluation formula for each pixel having the same coordinates in an image acquired from an inspection object has been proposed (Patent Document 2). According to this method, the tolerance is objectively set as a statistically meaningful value without depending on the ability of the worker.

JP 2001-175865 A JP 2005-265661 A

In the method described in Patent Document 2, it is necessary to register many non-defective products in order for the tolerance to have a statistically meaningful value. It takes a lot of time to register.
Therefore, the present invention improves the method described in Patent Document 2, and it is not necessary to select a good product in advance in order to objectively set the tolerance as a statistically meaningful value. It is an object of the present invention to provide an inspection program and an inspection apparatus that do not depend on the ability of the system.

In order to solve the problem, the inspection program of the present invention is:
In order to determine the quality of a plurality of objects having the same specification, the computer is caused to execute the following first step, storage step, provisional determination step, second statistical step, and main determination step.

  In the first statistical step, a provisional good pixel range is created by calculating characteristic statistics for each pixel having the same coordinates for a group of digital images individually acquired from a predetermined number of objects of the same specification. To do. The characteristic is a characteristic obtained from imaging data of an object, for example, an optical characteristic such as luminance and chromaticity, or a thermal characteristic such as temperature. In the storing step, the digital image group is stored in a memory. Statistics may be calculated for each image, or statistics may be calculated collectively from a group of images on the memory. The number of objects for calculating the statistic is ensured to exceed the number of statistically reliable samples (empirically about 50).

  In the temporary determination step, it is determined whether or not the characteristic belongs to the temporary good pixel range for each pixel with respect to each image in the stored digital image group. In a second statistical step, a set good pixel range is created by calculating a statistic of the characteristic for each pixel determined to belong to the provisional good pixel range. The set good pixel range is created based on the statistic calculated using only the pixels determined to belong to the provisional good pixel range, that is, the statistic calculated excluding the part of each object that is significantly inferior in quality. Therefore, it is close to the true good pixel range.

  In this determination step, it is determined whether or not the characteristic belongs to the set good pixel range for each pixel with respect to a digital image acquired from an object having the same specification to be inspected. Usually, after this, an object in which all the pixels are determined to belong to the set good pixel range is selected as a good product. This operation may be performed by a person or may be executed by a computer. Since the set good pixel range is close to the true good pixel range and is objectively determined from the statistics of the object regardless of the subjectivity of the operator, a highly reliable detection result can be obtained. Further, among the pixels determined not to belong to the set good pixel range, a plurality of adjacent pixels are grouped, and the feature amount of the entire group is checked with a reference amount, and an object including a group exceeding the reference amount is included. It is preferable to select defective products and other objects as non-defective products. This is because it is possible to prevent over-detection of detecting a defective product even if it is a non-defective product due to the presence of a small defect that should be allowed. “The characteristic amount of the entire group” means, for example, the total area (or total number) of pixels belonging to the group, the length of the long side of the rectangle circumscribing the group, and the like.

  A ring buffer is preferably used as the memory. During execution of the main determination step, each time a digital image of an object is acquired, the digital image is stored in a ring buffer, and the ratio of the objects selected as non-defective products in the main determination step is checked against a threshold value. The steps may be further executed by the computer. Thereby, when the ratio of the target object selected as a non-defective product falls below the threshold value, the first statistical step, the temporary determination step, and the second statistical step are performed again on the digital image group stored in the ring buffer. It can be executed to update the set good pixel range and maintain high detection capability and yield.

In order to solve the problems of the present invention, an inspection apparatus according to the present invention comprises:
A device for determining the quality of a plurality of objects of the same specification,
A first statistical means for calculating a provisional good pixel range by calculating a statistical quantity of characteristics for each pixel having the same coordinates for a group of digital images individually acquired from a predetermined number of objects of the same specification; ,
A memory capable of storing the digital image group;
Provisional determination means for determining whether the characteristic belongs to the provisional good pixel range for each pixel for each image in the digital image group;
A second statistical unit that calculates a statistical amount of the characteristic for each pixel belonging to the provisional good pixel range among the pixels in the digital image group, and creates a set good pixel range;
And a main judging unit for judging whether or not the characteristic belongs to the set good pixel range for each pixel with respect to the digital image acquired from the object.

  This inspection apparatus may be a combination of circuit elements and mechanical elements that execute the respective means, or may be a mode in which a part of the inspection apparatus cooperates with a software program. In addition, it is preferable that the image forming apparatus further includes a sorting unit that sorts an object for which all pixels are determined to belong to the set good pixel range as a good product. Instead of this, the selection unit groups a plurality of adjacent pixels among the pixels determined not to belong to the set good pixel range, collates the feature amount of the entire group with the reference amount, and determines the reference amount. It is particularly preferable that the object including the exceeding group is selected as a defective product and the other objects are selected as non-defective products.

  As described above, according to the present invention, the tolerance is objectively set as a statistically meaningful value without depending on the ability of the operator, and thus the inspection accuracy and reliability are excellent. In addition, since it is not necessary to select good products in advance, it is excellent in operational efficiency.

It is a block diagram which shows the hardware constitutions of the test | inspection apparatus of 1st embodiment. It is a block diagram which shows the system configuration | structure of the computer used for the inspection apparatus of the embodiment. It is a flowchart which shows operation | movement in the first half at the time of the test | inspection setting of the test | inspection apparatus of the embodiment. It is a flowchart which shows the detail of the statistics calculation operation | movement in the first half at the time of the inspection setting of the inspection apparatus of the embodiment. It is a figure explaining the setting method of the determination information of the same apparatus. It is a frequency distribution diagram which shows the relationship of each range. It is a flowchart which shows operation | movement in the first half at the time of the test | inspection setting of the same apparatus. It is a flowchart which shows the operation | movement at the time of the test | inspection of the same apparatus. It is a figure explaining the definition of the defective pixel applied to the inspection apparatus of 2nd embodiment.

Embodiment 1
A first embodiment of the present invention will be described with reference to the drawings. In FIG. 1, the inspection apparatus 100 includes a CCD camera 20 that takes an image of an inspection object W, a computer 30 connected to the camera 20, a positioning controller 40, and an XYθ table 50. The positioning controller 40 is connected to the computer 30 and converts the spatial orientation of the object W into a unique coordinate value based on position information detected by a position sensor (not shown) or the like.

  As shown in a block diagram in FIG. 2, the computer 30 has an interface with an image input unit 1 having an interface that enables connection with the camera 20, an input device such as a positioning controller 40, a keyboard (not shown), and a mouse. An operation unit 2, a central processing unit 3, a hard disk 4 and a RAM 5 are provided. The hard disk 4 stores an inspection program that causes the central processing unit 3 to execute predetermined arithmetic processing when an instruction or data is sent from the image input unit 1 or the operation unit 2. The RAM 5 temporarily records image data sent from the image input unit 1 and position information sent from the positioning controller 4.

  Specifically, a part of the RAM 5 is set in a FIFO ring buffer 6 capable of recording a maximum of n pieces of image data (n is an integer), and image data is recorded sequentially from the top address, and the final address is reached. Returning to the top address, new image data is overwritten on the already recorded image data. In the remaining part of the RAM 5, a reference image storage unit 7, a positioning information storage unit 8, a determination information storage unit 9, and a first statistics storage unit are stored so that the data processed by the central processing unit 3 is stored for each stage. 13, a second statistic storage unit 15 and a luminance value storage unit 22 are provided. The hard disk 4 includes a central processing unit 3 as a positioning processing unit 11, a first statistical processing unit 12, a temporary determination unit 14, a second statistical processing unit 16, a main determination unit 17, a selection unit 18, and a monitoring unit 19. Each function file is recorded.

  In the first half of the examination setting, as shown in the flowcharts of FIGS. 3 and 4, an arbitrary object W is imaged by the camera 20, an image is acquired from the image input unit 1, and the image is stored in the RAM 5 (S1). . Next, the image stored in the RAM 5 is stored as a reference image in the reference image storage unit 7 (S2). Then, positioning information such as a contour line in the reference image and a search range a as shown in FIG. 5 is set for the reference image and stored in the positioning information storage unit 8 (S3). Also, inspection determination information, for example, inspection range b, etc. is set for the reference image and stored in the determination information storage unit 9 (S4).

  Next, an image is similarly acquired for another object W having the same specification, and the image is stored in the ring buffer 6 (S9). The positioning processing unit 11 detects the position of the image stored in the ring buffer 6 in S9 with reference to the positioning information storage unit 8. Then, a positioning image converted so that the same part as the position of the reference image stored in the reference image storage unit 7 has the same coordinates (x, y) is acquired, and the luminance value g (x, y) for each pixel is obtained. Stored in the luminance value storage unit 22 (S10). Subsequently, in the first statistical processing unit 12, the luminance value g (x, y) for each pixel of the positioning image acquired in S <b> 10 is calculated according to Equation 1 and the sum S (x, y) and the square sum SS (x , Y) and the number of samples N (x, y) are calculated (S12).

[Formula 1]
S (x, y) = S (x, y) + g (x, y)
SS (x, y) = SS (x, y) + g (x, y) × g (x, y)
N (x, y) = N (x, y) +1

  Thereafter, S9 to S12 are individually repeated while sequentially sending yet another n-1 objects W having the same specifications (S5 to S7), and then the average value E1 of the luminance values for each pixel of the same coordinate according to Equation 2. (X, y) and standard deviation σ1 (x, y) are calculated and stored in the first statistic storage unit 13 (S8). The obtained value defines the provisional good pixel range in the frequency distribution as shown in FIG. 6 for each pixel.

[Formula 2]
E1 (x, y) = S (x, y) / N (x, y)
V1 (x, y) = (SS (x, y) −N (x, y) × E1 (x, y) × E1 (x, y)) / (N (x, y) −1)
σ1 (x, y) = sqrt (V1 (x, y)) (sqrt: calculation of square root)

  In the latter half of the inspection setting, a positioning image is obtained from the image stored in the ring buffer 6 by the above procedure as shown in the flowchart of FIG. 7 (S13), and the first statistic storage unit 13 is referred to for the positioning image. Then, the provisional determination unit 14 determines whether or not the luminance value g (x, y) for each pixel satisfies the condition of Expression 3 (S14), and extracts the luminance value of the pixel that satisfies the condition. In Equation 3, α1 may be about 5, and β1 may be about 2 to 3, for example.

[Formula 3]
E1 (x, y) -max {α1, β1 × σ1 (x, y)} <g (x, y)
And g (x, y) <E1 (x, y) + max {α1, β1 × σ1 (x, y)}

  Then, with respect to the extracted luminance value, the second statistical processing is performed on the luminance value sum S (x, y), the sum of squares SS (x, y), and the number of samples N (x, y) according to Equation 1 for each pixel having the same coordinates. By calculating in the unit 16 (S15 to S17), the average value E2 (x, y) and the standard deviation σ2 (x, y) are calculated for each pixel of the same coordinate as in the case of E1 and σ1. It stores in the 2nd statistics storage part 15 (S18).

  At the time of inspection, images acquired from the image input unit 1 for all objects having the same specifications are sequentially stored in the ring buffer 6 (S19), positioning images are acquired (S20), and the second statistic storage unit 15 is stored. With reference to this, the main determination unit 17 determines whether or not the luminance value g (x, y) of each pixel falls within the condition of Formula 4 (the set good pixel range) (S21).

[Formula 4]
E2 (x, y) -max {α2, β2 × σ2 (x, y)} <g (x, y)
And g (x, y) <E2 (x, y) + max {α2, β2 × σ2 (x, y)}

  Then, the selection unit 18 determines that the object determined that all the pixels satisfy the condition of Equation 4 as a non-defective product or the object determined that at least one pixel does not satisfy the condition as a defective product. And a defective product flag are assigned and sorted (S22 to S23) and displayed on the monitor. The monitoring unit 19 is activated simultaneously with the start of the inspection, and constantly monitors the ratio of the number of objects selected as non-defective products in the main inspection (S24 to S25). Then, at the same time as the ratio falls below the set value, the processing after S18 is executed using the image stored in the ring buffer 6 most recently, and the average value E2 (x, y stored in the second statistic storage unit 15 is executed. ) And standard deviation σ2 (x, y) and the like are updated (S26).

Embodiment 2
In the first embodiment, only one pixel does not satisfy the condition of Equation 4, and the object is selected as a defective product (S22 to S23). On the other hand, in the second embodiment, the selection unit 18 groups a plurality of pixels adjacent to each other among the pixels determined not to satisfy the condition of Expression 4, and the group feature amount is set as a reference amount. Match. Then, assuming that the feature amount is the total area of the pixels belonging to the group, the selection unit 18 assigns a non-defective flag and a defective flag to the target including the group exceeding the reference area as a defective product and the other target objects as non-defective products. Sort out.

  For example, when the reference area is set to 4 pixels, the groups G1 and G2 shown in FIG. 9 are not regarded as defects because the total areas are 4 pixels and 3 pixels, respectively. Therefore, no matter how many such pixels are interspersed, the object is not determined to be defective, and overdetection is prevented. On the other hand, since the total area of the group G3 is 11 pixels and exceeds 4 pixels, the group G3 is regarded as a defect and the object is determined as a defective product.

  The feature amount is not limited to the area, but may be the long side of the circumscribed rectangle or the roundness of the circumscribed circle, or a combination of these and the area. For example, when the feature amount is the length of a long side of a rectangle circumscribing the group and the reference amount is set to 2 (one side of one pixel is set to 1), the length of the long side of the group G2 is 3. Therefore, it is regarded as a defect. Group G3 is also regarded as a defect because the length of the long side of the circumscribed rectangle is 5, as indicated by a broken line. Accordingly, the objects including the groups G2 and G3 are determined as defective products. On the other hand, the group G4 is not regarded as a defect because the length of the long side is 2.

20 Camera 30 Computer 40 Positioning Controller 50 XYθ Table 100 Inspection Device

Claims (10)

To determine the quality of multiple objects of the same specification,
A first statistical step for creating a provisional good pixel range by calculating a statistical quantity of characteristics for each pixel having the same coordinates for a group of digital images individually acquired from a predetermined number of objects of the same specification; ,
Storing the digital image group in a memory;
A provisional determination step for determining whether or not the characteristic belongs to the provisional good pixel range for each pixel with respect to each image in the stored digital image group;
A second statistical step of calculating a statistic of the characteristic for each pixel determined to belong to the provisional good pixel range and creating a set good pixel range;
An inspection program for executing, on a digital image acquired from an object, a main determination step for determining whether or not the characteristic belongs to the set good pixel range for each pixel.   The inspection program according to claim 1, further comprising a selection step of selecting an object determined that all pixels belong to the set good pixel range as a non-defective product and other objects as defective products.   Further, among the pixels determined not to belong to the set good pixel range, a plurality of adjacent pixels are grouped, the feature amount of the entire group is compared with the reference amount, and an object including a group exceeding the reference amount is obtained. The inspection program according to claim 1, further comprising a sorting step for sorting defective products and other objects as non-defective products. The memory is a ring buffer;
During the execution of the main determination step, each time a digital image of an object is acquired, the digital image is stored in a ring buffer, and a monitoring step of checking the ratio of the objects selected as non-defective products with a threshold value is further performed on the computer. The inspection program according to claim 2 or 3 to be executed.   When the ratio of objects selected as non-defective products falls below the threshold, the first statistical step, the temporary determination step, and the second statistical step are executed again on the digital image group stored in the ring buffer. The inspection program according to claim 4, wherein the set good pixel range is updated.   The inspection program according to claim 1, wherein the characteristic is an optical or thermal characteristic. A device for determining the quality of a plurality of objects of the same specification,
A first statistical means for calculating a provisional good pixel range by calculating a statistical quantity of characteristics for each pixel having the same coordinates for a group of digital images individually acquired from a predetermined number of objects of the same specification; ,
A memory capable of storing the digital image group;
Provisional determination means for determining whether the characteristic belongs to the provisional good pixel range for each pixel for each image in the digital image group;
A second statistical unit that calculates a statistical amount of the characteristic for each pixel belonging to the provisional good pixel range among the pixels in the digital image group, and creates a set good pixel range;
An inspection apparatus comprising: a determination unit that determines, for each pixel, whether the characteristic belongs to the set good pixel range for a digital image acquired from an object.   The inspection apparatus according to claim 7, further comprising a sorting unit that sorts an object for which all pixels are determined to belong to the set good pixel range as a good product.   Further, among the pixels determined not to belong to the set good pixel range, a plurality of adjacent pixels are grouped, the feature amount of the entire group is compared with the reference amount, and an object including a group exceeding the reference amount is obtained. The inspection apparatus according to claim 7, further comprising a sorting unit that sorts defective products and other objects as non-defective products.   A recording medium storing the program according to claim 1.

Images