JP2002319021A - Binarization processing method, appearance inspection method and appearance inspection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately automatically binarize data having various distributions. SOLUTION: In a grouping process (S1-S5) of this binarization processing method, by applying a K-means method to the luminance data of an image, the luminance data are divided into a prescribed number of groups, and a representative value of each the group is calculated. In a binarization threshold value setting process (S6, S7), a binarization threshold is set in the middle point of a maximum distance section wherein a difference between the adjacent representative values in a luminance level is maximum. Thereby, even the image having an unknown area ratio between a background and a figure conventionally difficult to binarize, or the image having an unclear trough of the luminance distribution can be accurately automatically binarized. Accordingly, in an appearance inspection or the like for a chip type electronic component, the picked-up image can be analyzed to execute decision having little false recognition.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a binarization processing method in image processing and the like, and a visual inspection method and a visual inspection apparatus using the same.

[0002]

2. Description of the Related Art Conventionally, it has been practiced to irradiate an inspection object with light and detect defects of the inspection object with the naked eye or a visual inspection apparatus. The appearance inspection apparatus performs image processing on image data obtained by capturing the appearance of the inspection object,
Some automatically detect defects. For example, a chip-type electronic component appearance inspection apparatus can automatically detect defects such as cracks, chips, and adherence of foreign matter on the surface of a ceramic body by image processing using a luminance histogram.

[0003] Here, an appearance selection device (appearance inspection device),
In a positioning device, a character recognition device, and the like, a binarization process is performed in the process of image processing. As conventional binarization processing methods, there are a fixed threshold binarization method, a discriminant analysis method, a percent tile method, and the like.

[0004]

However, the conventional binarization processing method cannot accurately binarize an image in which the luminance level and the area ratio of the image and the background are not constant. For example, in the fixed threshold binarization method, since the binarization threshold is fixed, it can be applied only to an object whose image and background luminances are known in advance. The discriminant analysis method can be used for automatic binarization, but the threshold value cannot be set accurately unless the area ratio between the image and the background is close to some extent. The percent tile method cannot be accurately binarized unless the area ratio between the image and the background is known in advance.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to provide a binarization processing method capable of accurately and automatically binarizing data of various distributions. To provide a visual inspection method and a visual inspection device using the same.

[0006]

In order to solve the above-mentioned problems, a binarization processing method of the present invention performs a K-means method on an input data group to reduce a predetermined number of data of the data group. A K-means method using a K-means method processing means for dividing a data group consisting of a plurality of data that can be arranged one-dimensionally and acquired by a data acquisition means. Means for dividing into a predetermined number of groups, calculating a representative value of each group, and binarizing within a maximum distance section in which the difference between the adjacent representative values on the array is maximum. A threshold setting process for setting a threshold.

With the above configuration, in the data group grouping process, the data group is subjected to the K-means method, the data is divided into a predetermined number of groups, and a representative value of each group is calculated. In the threshold setting process, the binarization threshold of the data group is set in the maximum distance section in which the difference between the representative values adjacent in the array (for example, the luminance is high if the data group is an image) is the largest. Note that the binarization threshold may be the middle point of the maximum distance section on the histogram.

As a result, even if it is an image for which binarization has been difficult in the past, such as an image in which the area ratio between the image and the background is unknown or an image in which the valley of the luminance distribution is not clear, binarization can be accurately performed. Therefore, according to the above-described binarization processing method, it is possible to accurately and automatically binarize histograms having various shapes. Therefore, since the physical length such as the area, the center of gravity, and the perimeter of the image can be accurately measured, image processing such as highly accurate positioning, appearance selection, and character recognition can be performed.

In particular, when the above-described binarization processing method is applied to a visual inspection apparatus for chip-type electronic components, etc., it becomes possible to analyze a picked-up image which could not be analyzed conventionally, and it is possible to perform determination with less erroneous recognition. Therefore, accurate automatic binarization can be realized, and the accuracy of defect detection can be dramatically improved.

In order to solve the above-mentioned problems, an appearance inspection method according to the present invention includes an imaging process for imaging the appearance of an inspection object;
A binarized image creation process for binarizing luminance data of an image captured in the imaging process to create a binarized image; and a defect for determining a defect to be inspected based on the binarized image. And wherein the binarized image creating process performs a K-means method on the luminance data of the image, divides the luminance data into a predetermined number of groups, and represents a representative of each group. Grouping processing to calculate the value,
And a binarization threshold setting process of setting a binarization threshold in a maximum distance section in which the difference between the representative values adjacent to each other in the luminance level is maximum.

According to another aspect of the present invention, there is provided an appearance inspection apparatus which binarizes luminance data of an image photographed by the imaging unit, the photographing unit photographing the appearance of the inspection object. A binarized image creating unit that creates a binarized image; and a defect determining unit that determines a defect to be inspected based on the binarized image. Is K to the brightness data of the image.
A grouping unit that divides the luminance data into a predetermined number of groups by performing a means method and calculates a representative value of each group; and a maximum distance section in which the difference between the representative values adjacent to each other in the luminance level is maximum. And a binarization threshold setting means for setting a binarization threshold.

With the above arrangement, in the appearance inspection for judging the defect of the inspection object based on the binarized image obtained by binarizing the luminance data of the captured image of the appearance of the inspection object,
By performing the K-means method on the luminance data of the captured image, the pixels are divided into a predetermined number of groups, the representative value of each group is calculated, and the maximum distance at which the difference between the representative values whose luminance levels are adjacent is the maximum. A binarization threshold for luminance data is set in the section. Note that the binarization threshold may be the middle point of the maximum distance section on the luminance histogram.

[0013] With this, even if it is a captured image for which binarization has been difficult in the past, such as a captured image in which the area ratio between the defective portion and the normal portion is unknown, or a captured image in which the valley of the luminance distribution is not clear, it can be accurately binarized. Can be valued. Therefore, according to the binarization processing method, it is possible to automatically binarize luminance histograms of various shapes accurately.

Therefore, the appearance inspection method and the appearance inspection apparatus using the above-described binarization processing method for image processing can perform an appearance inspection which is robust against a defect state and a change in an imaging environment. . That is, it is hardly affected by a change in the posture of the target work, a change in the illumination, and the like, and there is little erroneous recognition. Further, since geometric information such as the area of the defect and the diameter of the Feret can be obtained, the defect can be determined based on the information.

Therefore, according to the above-described appearance inspection method and appearance inspection apparatus, a defective product having a defect does not flow to a subsequent process. In addition, since it is possible to know the size of the defect, information on the level of the defect that has occurred can be fed back to the manufacturing process to improve the quality of the product to be inspected. further,
It is possible to prevent an object having a shape and a size that is not defective from being excessively selected as a defect.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to FIGS.

In the present embodiment, regarding the binarization processing method according to the present invention, an external appearance in which an appearance inspection for a crack generated inside a ceramic body (inspection object) of a chip-type electronic component is automatically performed by image processing. The inspection apparatus 10 (FIG. 2) will be described as an example. It should be noted that the binarization processing method according to the present invention is not limited in its application range to image processing, but can be generally applied to binarization processing. In addition, the appearance inspection device 10 includes:
In addition to the internal cracks in the ceramic body, for example, the appearance of a scratch on the surface of the external electrode can be inspected by image processing using the binarization process.

FIG. 5A is an explanatory view showing a target work T to be inspected by the appearance inspection apparatus 10.

The target work T is a chip-type electronic component having a multilayer structure and usually a rectangular parallelepiped shape of 5 mm square or less. As shown in FIG.
Is composed of a ceramic body (ceramic body) T1 and electrodes (external electrodes) T2 formed at both ends thereof.

The ceramic body T1 is a multilayer structure mainly composed of ceramics and has a low light reflectance. On the other hand, the electrode portion T2 is a multilayer structure of metal and metal plating such as nickel and solder, and has a substantially smooth surface.
The light reflectance is high (it may be a mirror surface) and is almost constant.

Here, the following defects occur in the target work T as defects that can be detected by visual inspection. In the ceramic body T1, defects such as cracks and chips on the surface, adhesion of foreign matters such as ceramic chips, internal cracks and peeling of layers (internal cracks) occur. In the electrode portion T2, defects such as scratches that are irregularities on the surface, peeling of the electrode where the electrode floats and the base is exposed, and adhesion of foreign matter such as ceramic chips to the electrode surface occur.

FIG. 2 is an explanatory view showing the outline of the configuration of the appearance inspection apparatus 10, and shows a state viewed from above. FIG.
As shown in FIG.
Supply unit 12, imaging unit (data acquisition unit, imaging unit) 1
3. It comprises at least a sorting / discharging section 14 and a control section 15.

The mounting table 11 is a glass table having a disk-shaped mounting surface that rotates horizontally. Mounting table 11
, The target workpiece T placed with the surface to be inspected facing up,
Due to the rotation of the mounting surface, the sheet is conveyed to the imaging unit 13 and the sorting / discharging unit 14 which are sequentially provided on the periphery of the mounting surface.

The supply unit 12 is a linear feeder that supplies the target works T one by one onto the mounting surface of the mounting table 11.

The imaging unit 13 sets an inspection area S (FIG. 5A) on the transported target work T, captures an image, and transmits the acquired image data to the control unit 15. In addition, since the target work T moves at high speed by the mounting table 11, one imaging unit 13 performs imaging under one illumination condition. Therefore, when judging non-defective / defective products based on a plurality of imaging data, the plurality of imaging units 13 may be appropriately provided.

The sorting / discharging section 14 discharges the transported target work T in accordance with the result of the control section 15 determining a non-defective / defective product.

The control section 15 controls the entire appearance inspection apparatus 10. For example, based on the supply timing, the timing of imaging by the imaging unit 13 and the timing of discharge by the sorting / discharging unit 14 are managed. In addition, according to the above-described appearance inspection method, the non-defective / defective product of the target work T is determined.

Here, the control unit 15 can be configured based on a general-purpose computer. That is, the control unit 1
Reference numeral 5 denotes a CPU (central processing unit) for executing instructions of a program for realizing each function, a ROM (read only memory) storing boot logic, a RAM (random access memory) for expanding the program,
A storage device (recording medium) such as a hard disk for storing the above programs and various databases, input devices such as a keyboard and a mouse, output devices such as a monitor, a speaker and a printer, and network connection devices connected to an external network are connected by an internal bus. Connected and configured. The function of the control unit 15 is as follows.
, Respectively.

FIG. 3 is an explanatory diagram showing the outline of the configuration of the image pickup section 13. As shown in FIG. As shown in FIG.
Is equipped with a camera 41 and a ring illuminator 42 for capturing an image of the target work T placed on the mounting table 11 and being transported.

The camera 41 is used to move the target workpiece T being transported.
Is an imaging device that sets an inspection area S in the inspection area S and captures an image of the inspection area S.

The ring illuminator 42 is an illuminating device that illuminates the target work T with illumination light when an image is taken by the camera 41. The ring illuminator 42 is formed in a ring shape so that light from a light source can be uniformly illuminated from the surroundings of the target work T conveyed to a predetermined imaging position below the center of the ring illuminator 42 by an optical fiber. In particular, the ring illuminator 42 is disposed so as to emit illumination light from above the target work T at a preset illumination angle θ.

Here, the internal cracks in the ceramic body T1 are easily visible or hard to see depending on the illumination conditions such as the illumination angle and the intensity at the time of imaging. Therefore, the illumination condition of the imaging unit 13 is extracted and optimally set in advance by an experiment using a sample work (illumination condition setting processing). That is, the illumination condition is set such that the area of the defective portion D shown in FIG. 5A is large and the luminance difference between the defective portion D and the normal portion N is large.

FIG. 4 is a functional block diagram showing an outline of a defect detection and a non-defective / defective selection process in the visual inspection apparatus 10. As shown in FIG. The imaging unit 13 and the sorting / discharging unit 14 are the same as those shown in FIG. The image storage unit (captured image storage unit) 15A is provided in a memory or the like of the control unit 15. The binarized image creating unit (binary image creating unit) 15B and the defect determining unit (defect determining unit) 15C are realized by the control unit 15 executing a corresponding program.

The imaging section 13 sets an inspection area S in the ceramic body T1 of the target work T mounted and conveyed on the mounting table 11 (FIG. 5A), and the captured image data is controlled by the control section. 15 (imaging process). Next, the control unit 15 first stores the image data captured by the imaging unit 13 in the image storage unit 15A. Second, the binarized image creation unit 15
B obtains a binarization threshold from the luminance data (data group) of the image stored in the image storage unit 15A, and creates a binarized image of the image using the binarization threshold (binary image Creation process). Third, the defect determination unit 15C determines whether or not the target work T has a defect based on the binarized image created by the binarized image creation unit 15B (defect determination processing). Finally, the control unit 15 controls the sorting / discharging unit 14 so as to collect the target work T separately from non-defective products and defective products according to the judgment result of the defect judgment unit 15C.

Here, the binarized image creating unit 15B
Grouping processing unit (grouping means) 21, K-mea
An ns method processing unit (K-means method processing unit) 22 and a binarization threshold setting processing unit (binarization threshold setting unit) 23 are included. Then, the binarized image creating unit 15B converts the luminance data of the captured image read from the image storage unit 15A into
The grouping processing unit 21 divides the data into a predetermined number of groups according to the K-means method to obtain a representative value, and the binarization threshold setting processing unit 23 calculates the representative value in the maximum distance section where the difference between the adjacent representative values is maximum. A binarization threshold is set for a point, and a binarized image is created from the captured image using the obtained binarization threshold.

In the process of grouping the luminance data to obtain a representative value, the grouping processing section 21
The K-means method processing unit 22 is requested to perform the means method. Further, the process of grouping pixels into a predetermined number of luminances performed by the K-means method processing unit 22 corresponds to a region division process of dividing a luminance histogram into a predetermined number of regions. Further, the binarization threshold set by the binarization threshold setting processing unit 23 in the maximum distance section is not limited to the midpoint of the maximum distance section, and is determined according to various criteria such as the minimum value and the maximum curvature in the section. it can.

FIG. 7 is a graph showing the binarization threshold value TH of a certain luminance histogram according to the binarization processing method. In FIG. 7, assuming that the number of areas K is 3, the area is divided into three areas G1, G2, and G3. And
The distance between adjacent representative values (region average luminance) is L12
Since (= | R2-R1 |) <L23 (= | R3-R2 |), the maximum value is L23. Therefore, the binarization threshold TH is (R2 + R3) / 2. In FIG. 7, the binarization threshold TH is located at the middle point of the luminance histogram.

FIGS. 8 (a) and 8 (b) show the same luminance histogram as that of FIG. 7, with the number of areas K being 4 and 5, performing area division to extract the maximum distance section and obtaining the binarization threshold TH. FIG. As shown in FIGS. 7 and 8A and 8B, the representative values of the respective regions obtained by the K-means method are gathered at the background mountain and the image mountain. The number of representative values gathered in each mountain differs according to the area ratio between the image and the background.
In any case, the part where the difference between the adjacent representative values is large is the valley part of the luminance histogram. That is, the valley is always included in the maximum distance section.

Therefore, when a binarized image is created by setting a binarized threshold value at the valley portion, the image and the background can be clearly separated. Therefore, the physical length such as the area, the center of gravity, and the perimeter of the image can be accurately measured, and highly accurate positioning processing and appearance selection processing can be performed.

For example, FIG. 9 is a luminance histogram of the original image obtained by the experiment of FIG. 10A, in which the horizontal axis represents the luminance value and the vertical axis represents the number of pixels (frequency). From the luminance histogram, representative values 1 to 4 are obtained by setting the number of regions K to 4 by the K-means method. In FIG. 9, a section having both ends of the representative value 1 and the representative value 2 is the maximum distance section, and thus the binarization threshold TH can be obtained as an average value of this section. FIG. 10B shows a binarized image created by the binarization threshold TH, and a defective portion is extracted as a dark portion.

Next, with reference to the flowchart shown in FIG. 1, a process of converting a multi-tone image picked up by the image pickup unit 13 into a binarized image by the binarized image creating unit 15B will be described. Note that steps S1 to S5 are K-me
This is a process executed by the ans method processing unit 22 according to the K-means method, which is a method of dividing data into arbitrary regions.

First, in step S1, K-means
The number of regions (that is, the number of pixel groups)
Set K. The number of regions K is usually about 3 to 10, and 3 to 5 is suitable in the experimental results of the appearance inspection.
3 is optimal.

In step S2, the initial values of the representative values Y1 (n), Y2 (n),..., YK (n) are set in the K areas. Here, n indicates the number of repetitions. Therefore, the initial values of the representative values are Y1 (1), Y2 (1),.
K (1). In addition, since the representative value converges by repeated calculation, the initial value may be arbitrarily set. For example, a value obtained by equally dividing the luminance gradation by K + 1 may be set. That is, when an image of 256 gradations is divided into three luminance groups, the representative value is represented by luminance level 6
4, 128, and 192, respectively.

In step S3, assuming that the luminance of the pixel of interest (i, j) is aij, dm = | aij-Ym (n) | (m
= 1, 2,..., K) and the minimum value of dm dmin = m
Let m, which is in {dm}, be the region label of the pixel of interest (i, j). This calculation is performed for all the pixels, and each of the pixels is assigned one of the K region labels.

In step S4, the average value of the luminance of the pixels having the same region label (region average luminance) is calculated, and the representative value of each region is updated with the obtained region average luminance. Thereby, Y1 (2), Y2 (2),..., YK (2) are obtained.

In step S5, it is determined whether or not the region labels of all pixels are the same as the previous region label (FIG. 1). If not, the process returns to step S3. That is, the processing of steps S3 and S4 is repeated until the area labels of all the pixels converge.

As a result of the above processing (K-means method processing), assuming that the processing is completed q times, the final representative values (area average luminance) are Y1 (q), Y2 (q),.
(Q).

Subsequently, in step S6, the grouping processing section 21 sets the final representative values Y1 (q), Y2 (q),
.., YK (q) are rearranged in ascending order (or descending order), and a section (maximum distance section) of a representative value (average luminance) in which the difference between adjacent representative values is maximum is extracted. Next, in step S7, the binarization threshold setting processing unit 23 obtains the midpoint of the maximum distance section, and determines the obtained luminance as the binarization threshold TH of the captured image. Steps S6 and S7 correspond to a binarization threshold setting process.

Finally, in step S8, the binarized image creating section 15B creates a binarized image of the captured image by using the above-mentioned binarization threshold TH.

Here, a specific description will be given of a case in which the appearance inspection device 10 detects an internal crack generated in the ceramic body T1 by image processing using the above-described binarization method.

FIG. 5B is a luminance histogram of the inspection area S set in the ceramic body T1 having a defect of an internal crack. As shown in FIG. 5B, the internal crack of the ceramic body T1 appears as a high-luminance defect portion (high-luminance region) D as luminance data. That is, the internal cracks of the ceramic body T1
It looks brighter.

Here, the internal crack generated in the ceramic body T1 is an internal defect, and the surface thereof has a thickness of at most 1 μm.
Appears only as a float. Therefore, it is difficult to observe the internal crack as unevenness generated on the surface. However, when the surface of the ceramic body T1 is irradiated with light having an appropriate illumination angle θ (FIG. 3), the light permeates from the surface to the inside. Light that has entered the interior is reflected there if there is an internal crack, and exudes from the surface, whereas reflection does not occur at a portion without the internal crack and does not return to the surface. Therefore, by irradiating the surface of the ceramic body T1 with light having an appropriate illumination angle θ, the internal cracks can be observed as white areas. Note that the illumination angle θ
If it is high, it is reflected on the surface, so it is not known whether it is reflected inside. Further, the cracks or chips on the surface of the ceramic body T1 appear black without reflecting light regardless of the illumination angle θ.

Here, when detecting an internal crack in the ceramic body T1, the illumination angle θ is set to about 40 degrees. It is desirable that the lower limit value of the illumination angle θ is 30 degrees or more, preferably 35 degrees or more, and the upper limit value is 60 degrees or less, preferably 50 degrees or less. As described above, in the imaging unit 13, the optimal illumination condition for detecting the internal crack is set according to the target work T, so that the area of the defective portion D where the internal crack has occurred can be clearly imaged ( FIG. 5 (a)).

The appearance inspection apparatus 10 employs a determination algorithm utilizing the fact that the internal cracks of the ceramic body T1 appear as defective portions D having much higher brightness (brightness) than the brightness of the normal portion N. Used. That is,
The visual inspection device 10 sets the inspection area S in the ceramic body T1 by the imaging unit 13 and captures an image (imaging processing), and binarizes the brightness of the captured image by the binarized image creation unit 15B to obtain a binary image. High brightness area (defect portion D
(FIG. 6)) is calculated (binary image creation processing),
If the occupancy is equal to or more than the reference occupancy by the defect determination unit 15C, it is determined that the defective product has an internal crack (defect determination processing). Note that the reference occupancy is an allowable limit of the area of the defective portion D, and is a threshold value of the occupancy of the defective portion D when determining whether or not the defective portion is defective. Then, the reference occupancy is set in advance based on an experiment on the sample work.

Therefore, in the above-described visual inspection apparatus, in order to set the illumination condition optimally for detecting an internal crack according to the target work T, it is possible to clearly image a defect such as an internal crack and to automatically detect the defect. It becomes possible.

As described above, according to the above-mentioned binarization processing method, it is possible to binarize histograms of various shapes accurately. In other words, even if the area ratio between the image and the background of the image, which has conventionally been difficult to binarize, varies with the image, the binarization can be finely performed. For this reason, an image which could not be analyzed conventionally can be analyzed, and the image processing technology can be dramatically improved.

Therefore, the appearance inspection apparatus using the above-described binarization processing method for image processing can perform an appearance inspection which is robust against a defect state and a change in an imaging environment. That is, the appearance inspection apparatus is hardly affected by a change in the posture of the target work T, a change in the illumination, and the like.
Less misrecognition. Further, since geometric information such as the area of the defect and the diameter of the Feret can be obtained, the defect can be determined based on the information.

Therefore, according to the appearance inspection apparatus,
A defective product having a defect is not sent to a subsequent process. Also,
Since it is also possible to know the size of the defect, it is possible to improve the quality of the product to be inspected by feeding back information on the level of the defect that has occurred to the manufacturing process. Further, it is possible to prevent a material having a shape and a size that is not defective from being excessively selected as a defect.

The binarization processing method in the visual inspection apparatus 10 is a method of dividing a region into an arbitrary number of regions when binarizing a multi-tone density level sampled in pixel units. The means method is performed to determine a representative value (average luminance) of each area, arrange them in ascending order (descending order is also possible), determine the difference (absolute value) between adjacent representative values (average luminance), and determine the maximum difference. The binarization process may be performed by using the middle point between the representative values (average luminance) as the binarization threshold.

As a result, even if the area ratio between the image and the background of the image, which has conventionally been difficult to binarize, varies depending on the image, the binarization can be finely performed. For this reason, an image which could not be analyzed conventionally can be analyzed, and the image processing technology can be dramatically improved.

[0061]

According to the binarization processing method of the present invention, the input data group is subjected to the K-means method as described above,
A binarization processing method using K-means processing means for dividing the data of the data group into a predetermined number of groups, wherein the data group includes a plurality of one-dimensionally arrayable data acquired by the data acquisition means With the above K-mea
a grouping process of inputting to the ns method processing means, dividing the data into a predetermined number of groups, and calculating a representative value of each group;
And a binarization threshold setting process of setting a binarization threshold in a maximum distance section in which a difference between the representative values adjacent to each other on the array is maximum.

Therefore, it is possible to accurately binarize even an image for which binarization has conventionally been difficult, such as an image whose area ratio between an image and a background is unknown or an image in which a valley of a luminance distribution is not clear.
Therefore, according to the above-mentioned binarization processing method, there is an effect that histograms of various shapes can be accurately and automatically binarized. Therefore, since the physical length such as the area, the center of gravity, and the perimeter of the image can be accurately measured, there is an effect that image processing such as highly accurate positioning, appearance selection, and character recognition can be performed.

In particular, when the above-mentioned binarization processing method is applied to a visual inspection apparatus for chip-type electronic components, it is possible to analyze a captured image which could not be analyzed conventionally, and it is possible to make a determination with less erroneous recognition. Therefore, accurate automatic binarization can be realized, and the accuracy of defect detection can be dramatically improved.

As described above, the appearance inspection method of the present invention
An imaging process for imaging the appearance of the inspection target, a binarized image creation process for binarizing luminance data of the image captured in the imaging process, and creating a binarized image, based on the binarized image And a defect determination process of determining the defect to be inspected, and wherein the binarized image creation process performs a K-means method on the brightness data of the image to determine the brightness data. A grouping process of dividing the number of groups and calculating a representative value of each group, and a binarization threshold for setting a binarization threshold within a maximum distance section in which a difference between the representative values adjacent to each other in the luminance level is the maximum And a setting process.

Further, as described above, the appearance inspection apparatus of the present invention provides an image pickup means for picking up the appearance of an object to be inspected, and binarization of the luminance data of the image picked up by the image pickup means. And a defect judging means for judging a defect to be inspected based on the binarized image, and wherein the binarized image creating means comprises: K-means for the luminance data of
And a grouping means for dividing the luminance data into a predetermined number of groups and calculating a representative value of each group. And a binarization threshold setting means for setting a binarization threshold.

Therefore, even if the image ratio of the defect area to be inspected to the normal area is unknown, or even if it is an image in which binarization has been difficult in the past, such as an image in which the valley of the luminance distribution is not clear, It can be binarized accurately. Therefore, according to the binarization processing method, it is possible to automatically binarize luminance histograms of various shapes accurately.

Accordingly, an effect is obtained that an appearance inspection with high robustness can be performed even for a defect state and a change in an imaging environment. In other words, there is an effect that it is hardly affected by a change in the posture of the target work, a change in the illumination, or the like, and the erroneous recognition can be reduced. Further, since it is possible to obtain geometric information such as the area of the defect and the diameter of the feret, it is possible to determine the defect based on the information.

Therefore, there is an effect that a defective product having a defect does not flow to a subsequent process. Also, since it is possible to know the size of the defect, it is possible to improve the quality of the product to be inspected by feeding back information on the level of the defect occurring to the manufacturing process. It works. Further, there is an effect that it is possible to prevent an object having a shape and a size that is not defective from being excessively selected as a defect.

[Brief description of the drawings]

FIG. 1 is a flowchart schematically showing a binarization processing method according to an embodiment of the present invention.

FIG. 2 is an explanatory view schematically showing a configuration of a visual inspection apparatus in which the binarization processing method shown in FIG. 1 is applied to image processing.

FIG. 3 is an explanatory diagram schematically showing a configuration of an imaging unit included in the appearance inspection apparatus shown in FIG. 2;

FIG. 4 is a functional block diagram showing an outline of a sorting process in the visual inspection device shown in FIG. 2;

FIG. 5A is an explanatory view showing a chip-type electronic component to be inspected by the appearance inspection apparatus shown in FIG. 2; The chip-type electronic component is a defective product having an internal crack in the ceramic body. FIG. 5 (b) is the same as FIG.
3 is a luminance histogram of a captured image of a ceramic body of the chip-type electronic component shown in FIG.

FIG. 6 is a diagram showing a binarized image obtained by binarizing a captured image of the ceramic body of the chip-type electronic component shown in FIG. 5A using the binarization threshold of luminance shown in FIG. 5B; FIG.

FIG. 7 is an explanatory diagram showing a state in which the number of regions is set to 3 in the luminance histogram and a binarization threshold is set by the binarization processing method shown in FIG. 1;

8 (a) and 8 (b) are explanatory diagrams of the comparative example of FIG. 7; FIG. 8 (a) shows a case where the number of regions is 4, FIG. 8 (b) shows a case where the number of regions is 5 Is shown.

9 is a specific example of the binarization processing in the appearance inspection device shown in FIG. 2, and is an explanatory diagram showing a luminance histogram and a binarization threshold of a captured image shown in FIG.

10A and 10B are specific examples of the binarization processing in the appearance inspection device shown in FIG. 2; FIG. 10A is a drawing substitute photograph showing a halftone image in which a captured image captured by an imaging unit is displayed on a display; FIG. 10B is a diagram showing a binarized image obtained by binarizing the captured image of FIG. 10A using the binarization threshold of FIG.

[Explanation of symbols]

Reference Signs List 10 appearance inspection device 13 imaging unit (data acquisition unit, imaging unit) 15B binarized image creation unit (binary image creation unit) 15C defect determination unit (defect determination unit) 21 grouping processing unit (grouping unit) 22 K-means method processing unit (K-means method processing means) 23 binarization threshold setting processing unit (binarization threshold setting means) S1 to S5 grouping processing S6, S7 binarization threshold setting processing T target work (inspection Target) TH binarization threshold

Continued on front page F-term (reference) 2G051 AA61 AB02 CA04 DA08 EA08 EA11 EB01 EC02 5B057 AA03 BA02 CA08 CA12 CA16 CB06 CB12 CB16 CC01 DA03 DB02 DB08 DC04 DC23 5C077 LL19 MP01 RR14

Claims (3)

[Claims] 1. A binarization processing method using K-means processing means for subjecting an input data group to a K-means method and dividing the data of the data group into a predetermined number of groups, A data group consisting of a plurality of data that can be arranged one-dimensionally and acquired by the data acquisition means is input to the K-means method processing means, divided into a predetermined number of groups, and a group for calculating a representative value of each group And a binarization threshold setting process of setting a binarization threshold within a maximum distance section in which a difference between the representative values adjacent on the array is maximum. Method. 2. An image pickup process for picking up an appearance of an inspection object, a binarized image forming process for binarizing luminance data of an image picked up by the image pickup process to generate a binarized image, A defect determination process for determining a defect to be inspected based on the binarized image, and the binarized image creation process performs a K-means method on luminance data of the image, Dividing the luminance data into a predetermined number of groups, and calculating a representative value of each group; and setting a binarization threshold value in a maximum distance section in which a difference between the representative values adjacent to each other in the luminance level is maximum. And a binarization threshold setting process. 3. An image pickup means for picking up an appearance of an inspection object, a binarized image forming means for binarizing luminance data of an image picked up by the image pickup means to generate a binarized image, Defect determination means for determining the defect to be inspected based on the digitized image, and wherein the binarized image creation means performs a K-means method on the luminance data of the image. Grouping means for dividing the luminance data into a predetermined number of groups and calculating a representative value of each group; and a binarization threshold value within a maximum distance section in which the difference between the representative values adjacent to each other in the luminance level is maximum. And a binarization threshold setting means for setting.