JP2001034762A - Method and device for image processing checking - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing checking method and an image processing checking device which can extract an accurate edge element even though much pepper and salt noise caused by surface characteristic is included in an image and also an object to be checked has small variable-density difference of a defected image. SOLUTION: This image processing checking device generates a binarized image by a density gradient calculating part 21 calculating the density gradient of each pixel, an average value calculating part 22 calculating the average value of density gradients of each pixel, a standard deviation calculating part 23 calculating the standard deviation of the density gradients of each pixel, a threshold calculating part 24 deciding the threshold T with which an edge where the density of image data changes is discriminated on the basis of the average value and the standard deviation, a binarization processing part 25 which binarizes image data on the basis of the threshold T and extracts an edge element including a defect, and a filtering part 26 which screens the binarized image data with a mesh-shaped filter having a plurality of square lattices in a prescribed size and eliminates noise components included in the edge element.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing inspection method and an image processing inspection apparatus for extracting a defect present on a surface of an object to be inspected, for example.

[0002]

2. Description of the Related Art For example, a sealing member such as an O-ring is
Before the product is shipped, it is necessary to inspect the sealing surface for defects such as scratches in order to exert the original function of the sealing member. Conventionally, the inspection of the seal surface of the O-ring has been performed by an inspection device as shown in FIG. FIG. 17A is a top view of the inspection apparatus, and FIG. 17B is a side view. In FIG. 17, an O-ring 101 is mounted on a rotating shaft 105a of a drive motor 105, and an image of the inspection surface 101a is taken on the side of the inspection surface 101a of the O-ring 101, for example, a CCD using a charge-coupled device. The camera 102 is arranged. Also, C
On both sides of the CD camera 102, illumination devices 103 for irradiating illumination light toward the inspection surface 101a of the O-ring 101 are arranged. Image data captured by the CCD camera 102 is input to the image processing device 106.

In the inspection apparatus having the above configuration, an image of the inspection surface 101a of the O-ring 101 illuminated with illumination light from the illumination device 103 is captured by the CCD camera 102, and the image is input to the image processing device 106. . The image input to the image processing device 106 is an image having a simple gray-scale structure. The image processing device 106 extracts, for example, target portions having different density values from the image,
After converting to a binarized image in which the value 1 is assigned to the target portion and the value 0 is assigned to the other portions to determine the edge element, the connection of each pixel of the edge element is checked. Different labels are assigned (labeling), and the shape and correlation distance of each label are analyzed to finally recognize the defect existing on the inspection surface 101a of the O-ring 101. Note that pixel concatenation means that two pixels a and b having the same value in a binary image have a sequence of adjacent (four or eight) pixels having the same value. It refers to the relationship between two pixels a and b. In the extraction of the edge element in the image processing device 106, a threshold value is determined, and an edge element whose density value changes is extracted from the image based on the threshold value. In addition, the edge elements extracted by the binarization processing are often line figures having a width alone, or are often divided into small line segments due to the influence of noise or the like.
It is difficult to recognize the shape of the defect as it is. For this reason, after the image processing device 106 extracts the edge element by the binarization processing of the image data, the image processing apparatus 106 performs an edge tracing processing of tracing a pixel having the same property among the pixels in the edge element. The connected trajectories are extracted to identify defects.

[0004]

The O-ring 101, which is the object to be inspected, is made of, for example, a rubber material. In general, rubber products contain a large amount of additives, so that their particles are not completely dispersed and the surface has non-uniform light reflection characteristics. In addition, the surface is often rough even at a level that does not cause a problem in performance due to the influence of the mold and raw materials of the O-ring 101. Further, there is a case where micro-level dust which has no problem in performance but cannot be removed by means such as blowing air is attached.
When the O-ring 101 having such surface characteristics is captured by the CCD camera 102, the captured image is an image including much noise due to the surface characteristics of the rubber product. Note that the noise here is so-called sesame salt noise, which is caused by the fact that an image in which the illuminance of the background is uniform and the defect is black cannot be obtained. O-ring 10
1 is, for example, the O-ring 101
Foreign matter adhering to the surface of the mold for molding the O-ring 1
01, or occurs due to improper filling of the molding material into the mold. For this reason, the state of the defect is also various, and in the image containing a lot of sesame salt noise, the density of the defect is not clear, that is, the density difference between the defect and the background is small, and the shape of the defect is accurate. Often cannot be recognized. In addition, sesame salt noise has two
This becomes remarkable when the value is converted into a value, and the image becomes an image in which black dots are scattered in a white region (background or non-defective surface) other than the defective region. On the other hand, as a method of determining a threshold necessary for performing binarization processing on image data, for example, a P tile method is known. The P tile method is a method of determining a threshold value such that the ratio between the area of an edge element including a defect area to be extracted and the area of a target image is constant. However, it is difficult to estimate the area of a defect existing on the surface of the O-ring 101 as described above in advance, and thus it is difficult to determine an appropriate threshold value. As other threshold value determination methods, a density histogram method and a density difference histogram method are known. In these methods, too, a threshold value for extracting an edge element from an image containing much noise is set. Difficult to decide. If the threshold value is not appropriate, the edge element cannot be extracted with high accuracy, and it is difficult to specify the defect with high accuracy in the subsequent processing.

Conventionally, binarization processing of image data has been performed after removing noise of an image containing much sesame salt noise in advance. As a noise removing method, for example, a local averaging method for averaging the densities of a target pixel and eight pixels adjacent thereto or a density histogram in a local region is obtained.
There are known a median filter having an intermediate value as an output value, a two-dimensional Fourier transform for performing Fourier transform of image data to remove high-frequency components, and performing an inverse Fourier transform of the data after removing the high-frequency components. However, when these noise removal methods are used, the processing time required for the noise removal becomes long, the time required for the surface inspection of the O-ring 101 becomes long, and there is a disadvantage that a practical inspection cannot be performed.

The present invention has been made in view of the above-described problems. For example, as in the case of an image of a defect existing on the surface of a rubber material, a large amount of sesame salt noise is contained in a region other than the defect region of the image. It is an object of the present invention to provide an image processing inspection method and an image processing inspection apparatus capable of detecting edge elements with high accuracy in binary image processing even when an inspection target has a small density difference of a defect image.

[0007]

According to an image processing and inspection method of the present invention, image data of the surface of an object to be inspected is binarized, and the surface of the object to be inspected is processed based on the binarized image data. An image processing inspection method for extracting a defect existing in the inspection object, wherein a step of obtaining a density gradient of each pixel of the image data of the surface of the inspection object and a step of obtaining an average value and a standard deviation of the density gradient of each pixel Determining a threshold value for determining an edge at which the density of the image data changes based on the average value and the standard deviation; and binarizing the image data based on the threshold value. Extracting an edge element containing a defect;
Screening the valued image data with a mesh filter having a plurality of square lattices of a predetermined size to remove noise components included in the edge elements, and the screened image data Is used to extract defects present on the surface of the inspection object.

The image processing / inspection apparatus of the present invention binarizes image data of the surface of the object to be inspected, and extracts a defect existing on the surface of the object to be inspected based on the binarized image data. An image processing and inspection apparatus, wherein a density gradient calculator for calculating a density gradient of each pixel of the image data of the surface of the inspection object, an average value calculator for calculating an average value of the density gradient of each pixel, A standard deviation calculating unit for calculating a standard deviation of a density gradient of each pixel; and a threshold calculating unit for determining a threshold value for determining an edge where a density of the image data changes based on the average value and the standard deviation. A binarization processing unit that binarizes the image data based on the threshold value and extracts an edge element including a defect; and converts the binarized image data into a plurality of square grids of a predetermined size. With a mesh filter Clean anda filter unit for removing noise components contained in the edge element,
A defect existing on the surface of the inspection object is extracted based on the screened image data.

According to the present invention, when binarizing image data of the surface of the inspection object to extract an edge element which is contour information of a defect existing on the surface of the inspection object, the average value of the density gradient of each pixel is extracted. And a standard deviation, the image data is binarized based on the threshold value.
An edge element which is a set of points at which the density of a pixel changes is extracted. By determining the threshold value from the average value and the standard deviation, the threshold value becomes an appropriate value even if the shape of the defect is not estimated in advance. The binarized image data contains a lot of sesame salt noise if the surface of the inspection object has a surface characteristic such as a rubber material, for example. By screening the binarized image data containing much noise with a mesh filter having a square lattice, that is, by sieving, the noise component is largely removed by the square lattice, and the edge element from which the noise has been removed is removed. Is obtained. The edge element reflects the contour shape of the defect existing on the surface of the inspection object with relatively high accuracy, and the defect can be specified with high accuracy in the subsequent defect extraction processing.

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing one configuration example of a surface inspection apparatus to which an image processing inspection apparatus according to the present invention is applied. As shown in FIG. 1, the surface inspection device 61 according to the present embodiment includes a driving device 64 that is fixed on a base 62 and that rotationally drives a mounting shaft 65 on which a seal member 81 as an inspection object is mounted. A CCD camera 71 as imaging means supported by a support member 63 provided on the base 62; a ring illumination 67 as illumination means fixed to the support member 63 and provided above the seal member 81; Roll type diffusion plate 66 provided above member 81
, The light source 70 of the ring illumination 67, and the image processing inspection device 7
And 5.

FIG. 2 shows a seal member 81 for a gas valve as an object to be inspected by the surface inspection apparatus 61 shown in FIG.
FIG. 3 is a cross-sectional view showing the structure of FIG. The seal member 81 has, for example, a through hole 81 at the center formed from a black rubber material.
b, and has an annular sealing surface 81a on one end surface side. The annular sealing surface 81a is formed such that the cross-sectional contour is curved in a convex shape, and the sealing surface 81a is in close contact with another member to perform a sealing function. If a defect such as a scratch is present on the sealing surface 81a of the sealing member 81, the sealing surface 81a does not perform its original function, so it is necessary to inspect for a defect.

The ring illumination 67 is an annular body having an opening 67a in a central region thereof, and incorporates a bundle of optical fibers arranged in an annular shape (not shown). These optical fibers are led out from the side of the ring illumination 67 toward the light source. The light supplied from the light source 70 is emitted as annular illumination light downward from the lower surface 67b of the ring illumination 67 through an optical fiber arranged annularly inside the ring illumination 67. The ring illumination 67 is preferably arranged such that the center of the opening 67 a is located on the sealing surface 81 a of the sealing member 81. In addition, ring illumination 67
The diameter of the opening 67a is set sufficiently larger than the radial width of the sealing surface 81a. The light source 70 includes
For example, a halogen lamp compatible with the CCD camera 71 as an image pickup means with respect to wavelength characteristics can be used.

The roll diffusion plate 66 is a cylindrical body having an opening 66a at the center. Roll type diffuser 66
The center of the opening 66a is the opening 67 of the ring illumination 67.
The ring light 67 is disposed so as to substantially coincide with the center line of a, and a part on the upper end side is inserted into the opening 67 a of the ring illumination 67. Further, the roll-type diffusion plate 66 has an opening 66.
The center of “a” is disposed so as to be located substantially at the center of the seal surface 81 a of the seal member 81. Roll diffuser 6
6 guides the annular illumination light emitted from the ring illumination 67 toward the outer peripheral surface 66c from the lower end surface 66b to the sealing surface 81a of the sealing member 81 through the annular solid portion. At this time, the annular lower end surface 66 of the roll diffusion plate 66
The illumination light diffused from b is diffused in various directions. As a material for forming the roll-type diffusion plate 66, for example, a translucent acrylic resin colored milky white or the like can be used, and this material is formed into a cylindrical shape. In addition, the diameter of the roll-type diffusion plate 66 is set to a value that can ensure the visual field range of the CCD camera 71 and the inspection range of the seal surface 81a of the seal member 81.

The driving device 64 includes, for example, the sealing member 8
1 comprises a drive motor for rotating the mounting shaft 65 at a predetermined angle, and sequentially rotates the seal member 81 at a rotation angle corresponding to each inspection area of the seal surface 81a of the seal member 81 imaged by the CCD camera 71. The seal member 81 is finally rotated 360 degrees.

The CCD camera 71 is arranged so that the field of view is directed to the sealing surface 81a of the sealing member 81 through the opening 66a of the roll diffusion plate 66. CCD camera 7
For 1, a camera generally used in the field of image processing is used, for example, a 1 million pixel CCD camera is used.

The image processing inspection device 75 is composed of, for example, a personal computer, receives image data picked up by the CCD camera 71, quantizes the image data, and performs image processing described later on the image data. The defect existing on the sealing surface 81a of the sealing member 81 is extracted.

In the surface inspection apparatus 61 having the above configuration, first,
The through-hole 81b of the seal member 81 is mounted on the mounting shaft 65 of the driving device 64. Next, illumination light is emitted from the ring illumination 67. Part of the annular illumination light emitted from the ring illumination 67 enters the outer peripheral surface 66c of the roll diffuser 66,
The light introduced into the roll-type diffusion plate 66 is diffused from the lower end surface 66b of the roll-type diffusion plate 66, and the inspection area of the seal surface 81a of the seal member 81 is irradiated with diffused light scattered from the lower end surface 66b. Therefore, the illuminance distribution in the inspection area of the seal surface 81a of the seal member 81 is determined by the
Even if 1a is curved, it becomes substantially uniform. In this state, an image of the inspection area is captured by the CCD camera 71 provided above the inspection area on the seal surface 81a of the seal member 81.

FIG. 3 is a configuration diagram showing the configuration of the image processing inspection apparatus 75 described above. As shown in FIG.
The image processing inspection device 75 is roughly divided into a binarization processing unit 2
0 and a defect extraction unit 30. The binarization processing unit 20
The image data of the seal surface 81a of the seal member 81 captured by the CCD camera 71 is binarized. The defect extraction unit 30 extracts a defect existing on the sealing surface 81a of the sealing member 81 based on the edge element of the image data binarized by the binarization processing unit 20.

The binarization processing section 20 includes an image data input section 2
1, a density gradient calculator 22, an average calculator 23, a standard deviation calculator 24, a threshold calculator 25, an edge element extractor 26, and a filter 27.

The image data input section 21 is provided with the CCD camera 7
1. Seal surface 81a of seal member 81 imaged by 1
And quantizes it.

The density gradient calculator 22 calculates the density gradient (gradient intensity) Δ of each pixel of the quantized image data input from the image data input unit 21. Specifically, a difference value of the density of each pixel is calculated.

The average value calculating section 23 includes a density gradient calculating section 22.
The average value m of the density gradient Δ of each pixel of the image data calculated by the above is obtained. The standard deviation calculation unit 24 includes the concentration gradient calculation unit 2
Find the standard deviation σ of the density gradient Δ of each pixel of the image data calculated in step 2.

The threshold value calculating section 24 includes an average value calculating section 23
Based on the average value m and the standard deviation σ of the density gradient Δ calculated in the above, a threshold value T for determining a changing point (edge element) where the density of the image data changes is determined. Threshold T
Is determined, for example, by the following equation (1).

T = m + t · σ (1) where t is a parameter for adjusting the threshold value T and is set as appropriate.

The edge element extraction unit 25 determines the seal surface 81a of the seal member 81 input to the image data input unit 21 based on the threshold value T calculated by the threshold value calculation unit 24.
Is binarized to extract an edge element including a defect. Specifically, the density of each pixel input to the image data input unit 21 is compared with a threshold value T, and a pixel having a density value larger than the threshold value T is determined as a defective pixel as an edge element. A pixel having a density smaller than the threshold value T is set as a background element. The edge element of the image data binarized by the threshold value T includes the sealing surface 8 of the sealing member 81.
1a and sesame salt noise components due to roughness of the sealing surface 81a and the like.

The filter unit 26 includes an edge element extraction unit 25
The binarized image data is screened by a mesh filter having a plurality of square grids of a predetermined size to remove sesame salt noise components included in edge elements.
Here, FIG. 4 is a diagram for explaining a process performed on the image data binarized by the filter unit 26. As shown in FIG. 4A, a rectangular pattern image R ₀ and a pattern image R
Assuming a binarized image in which a minute noise image N existing around ₀ exists as an edge element, the filter unit 2
6 screens, ie, screens, the binarized image data with a filter K having a plurality of square lattices of a predetermined size (hereinafter, referred to as a lattice filter K). In the filter unit 26, pixels that exist only in the area surrounded by the grid line KL of the grid filter K and are not present on the grid line KL in the pattern image _R0 and the noise image N are excluded from the edge elements. 2 with the above processing
The binarized image is a pattern image R as shown in FIG.
Only ₀ , and the noise image N is excluded. Note that the size of the square grid of the grid filter K is appropriately set in accordance with the size of noise included in the binarized image data.

In the binarization processing section 20 of the image processing / inspection apparatus 75 of this embodiment, the threshold value T is set to the average value m of the density gradient Δ.
And the standard deviation σ, as described above,
Even if the image of the seal surface 81a of the seal member 81 contains much noise and the defective image is buried in the noise, it is possible to easily determine the threshold value T of an appropriate value to some extent. In the binarization processing unit 20 of the image processing inspection device 75 of the present embodiment, binarization is performed by sieving the image data binarized based on the threshold value T using a lattice filter K. Noise components included in image data can be easily removed. As described above, in the binarization processing unit 20 of the image processing inspection apparatus 75 of the present embodiment, the threshold value T required for the binarization processing of the image data of the sealing surface 81a of the sealing member 81 is relatively simple. In addition, since a lattice filter K that does not require a complicated algorithm is used to remove noise components contained in binarized image data, the processing time required for noise removal is greatly reduced. can do. The edge element composed of the defective pixel of the binarized image data obtained in this manner becomes an image appropriately including information on the defect existing on the sealing surface 81a of the sealing member 81.

The defect tracing unit 30 includes the edge tracing unit 3
1. It has an inter-pixel connection investigation unit 32 and a labeling processing unit 33. The edge tracking processing unit 31 performs an edge tracking process of an edge element of the binarized image data obtained by the filter unit 26 of the binarization processing unit 20. Noise components are removed to some extent from the edge elements of the binarized image data obtained by the filter unit 26 of the binarization processing unit 20, but in the edge elements in this state, the noise components included in the image data are removed. In many cases, the edge elements are divided into small line segments due to the complexity of the defect and the shape of the defect, and the defect shape cannot be identified unless the broken edge elements are connected and smoothed. For this reason, the edge tracking processing unit 31 performs connection between edge elements by, for example, the method shown in FIG.

As shown in FIG. 5A, first, a local point (defect position) R1 of the edge element S reflecting a series of defects existing in the binarized image is determined. Next, a point r located on the circumference of the radius r centered on the local point R1
Among the pixels located on the line segment between [1] and the local point R1, the number of pixels having the same property as the pixel located at the local point R1 is counted. Specifically, the pixel having the same property is, for example, a pixel determined to be a defective pixel similarly to the pixel located at the local point R1. The same processing is performed at points r [2] to r [n] at positions rotated by an angle θ around the local point R1 on the circumference of the radius r (where r [n] is a position rotated by 180 degrees). ) And the local point R1, the number of pixels having the same properties as the local point R1 among the pixels located on the line segment is sequentially counted.

As a result of the above processing, if the direction in which the number of pixels having the same properties as the pixel located at the local point R1 becomes the maximum is the direction of the point r [a], the position at the local point R1 in this direction is determined. maximum number V _max of pixels having similar properties to pixels only if greater than a predetermined threshold value Th, as shown in FIG. 5 (c), this point a new local point R2 . Similarly, with the local point R2 as a starting point, the processing described in FIG.
As shown in FIG. 5D, new local points R3 and R4
To explore. For example, when a pixel having the same property as the pixel located at the local point R4 is searched starting from the new local point R4, the number of pixels having the same property as the pixel located at the local point R4 becomes the maximum. maximum number V _max of the pixel is smaller than the threshold value Th is the local point R4 and end point.

The local points R1, R thus searched for
A trajectory connecting 2, R3 and R4 is defined as a trajectory of one defect. By performing the above processing for each pixel of the edge element, the entire trajectory of the defect is specified.

The inter-pixel connection checking unit 32 checks the connectivity of each pixel present on the trajectory of the defect obtained by the edge tracking processing unit 31 based on the edge element of the binarized image.
For example, as shown in FIG. 6, a pixel x located at (i, j)
_{For 0} , (i + 1, j), (i, j + 1), (i-1, j), (i, j-1)
, (I + 1, j + 1), (i-1, j + 1), (i + 1, j-1), (i + 1, j-1)
Each pixel x ₁ ~x ₈ located is called the 8-neighbor, a pixel x ₀ the same value as the pixel of the eight neighboring that is adjacent to the pixel x ₀ in. In addition, for two pixels a and b having the same value in the binary image, two pixels a and b when there is a series of adjacent (four or eight) pixels having the same value are present. A relationship. The inter-pixel connection checking unit 32 performs a process of searching for a pixel connected to each pixel existing on the trajectory of the defect obtained by the edge tracking processing unit 31 among the edge elements as described above. In other words, the edge tracing processing unit 31 determines the global trajectory of the defect by globally connecting edge elements that are often broken into small line segments, but the actual shape of the defect is determined only by this trajectory. Since the pixel cannot be specified, a pixel connected to each pixel located on the trajectory is searched, and a defect is specified with high precision by expanding the defective pixel.

The labeling processing unit 33 assigns different labels to the set of connected pixels (connected components) obtained in the edge tracking processing unit 31 and the inter-pixel connection checking unit 32, and assigns different labels to the connected components in each label. By analyzing the shape and the correlation distance, the seal surface 8 of the seal member 81 is analyzed.
The final recognition of the defect existing in 1a is performed.

Next, an example of the processing procedure of the image processing inspection method using the image processing inspection apparatus will be described with reference to FIGS.
This will be described with reference to the flowchart shown in FIG. First, FIG.
As shown in FIG.
Is input to the image processing inspection device 75, and the density gradient Δ of each pixel of the image data is obtained (step S1). FIG. 9 shows an example of an image of the seal surface 81a of the seal member 81 input to the image processing inspection device 75. As shown in FIG. 9, the image of the seal surface 81 a includes a large amount of noise as a whole, and a defect on the seal surface 81 a having a high density is faintly displayed in the central region.

Next, the average value m and the standard deviation σ of the obtained density gradient Δ of each pixel are obtained (step S2), and the threshold value T is obtained from the average value m and the standard deviation σ by the above equation (1) (step S2). Step S3). Binarization of all pixels of the image data is performed using the threshold value T (step S
4). In this binarization, for example, a defective pixel (a pixel whose density is higher than a threshold value) is set to 1 and other pixels are set to 0. FIG. 10 shows an example of a binarized image obtained by binarizing the image shown in FIG. In FIG. 10, a black area is an area where the pixel value is 0, and a white area is an area where the value of the pixel is 1. As can be seen from FIG. 10, the binarized image contains many noise components.

Next, the above-mentioned lattice filter K is set for the target pixel of the above-mentioned binary image (step S).
5) It is determined whether there is a defective pixel on the grid line KL (step S6), and the image data is screened. If there is no defective pixel on the grid line KL, the value of the pixel in the grid surrounded by the grid line KL is set to 0 (step S7). If a defective pixel exists on the grid line KL, the value of the pixel in the grid surrounded by the grid line KL is left as it is. It is determined whether the grid filter K has been set for all the pixels (step S8). When the setting of the grid filter K is completed, a binarized image from which noise has been removed is obtained (step S9). Here, FIG.
FIG. 12 is a diagram illustrating a state where a lattice filter K is set in the binarized image, and FIG. 12 is a diagram illustrating a binarized image screened by the lattice filter K. Comparing the above-described binarized image shown in FIG. 10 with the binarized image shown in FIG. 12, it can be seen that most of the noise components have been removed. The processing up to this point is the processing performed in the binarization processing unit 20 of the image processing inspection device 75 described above.

Next, as shown in FIG. 8, an arbitrary defective pixel (pixel having a value of 1) is determined from the obtained binary image data, and this point is set as a local point L (step S1).
1). Next, the number M of defective pixels existing on a line segment connecting the point r [j] (j = 1 to n) on the circumference of the radius r from the local point L and the local point L is counted. Note that the point r [j] is at an angle Φ [i] (Φ = 0) from the reference position on the circumference.
１８０180 °). Based on the number M of defective pixels on a line segment connecting the local point L and each point r [j] on the circumference of the radius r, a direction Φ in which the number M of defective pixels is maximum
Search for max (step S12). The point on the circumference of the radius r and the local point L in the direction Φmax at which the number M of defective pixels is the maximum is defined as r [a].

Next, the number M of defective pixels in the direction Φmax in which the number M of defective pixels is maximized is compared with a predetermined threshold Th (Step S13), and the number M of defective pixels is determined by a threshold value. If it is greater than the value Th, the point r
[A] is set as a new local point L (step S14),
Starting from the new local point L, a direction Φmax in which the number M of defective pixels is maximized is searched. If the number M of defective pixels is smaller than the threshold value Th, the local point L is not updated, the local point L is set as the end point, and a locus connecting the local points L obtained by the search is determined ( Step S1
5).

FIG. 13 is a diagram showing a trajectory obtained by the defect tracing process performed on the binarized image shown in FIG. On the other hand, FIG. 15 is a diagram showing a trajectory obtained as a result of performing a tracking process of the defect trajectory by examining the eight-point connection of each pixel of the binary image shown in FIG. The tracking of the defect trajectory by connecting the eight points of each pixel of the binarized image is a local tracking, and as can be seen from FIG. 15, the ratio of the tracked trajectory is divided into small line segments. On the other hand, according to the tracking method of the present embodiment, as shown in FIG. 13, the ratio of the trajectory divided into small line segments is small, and the feature of the trajectory of the defect can be grasped globally.

Next, the connectivity of each pixel on the obtained defect trajectory is examined (step S16). The eight-point connection of each pixel on the global defect trajectory obtained by the above processing is examined, and the defective pixel is extended by obtaining a connected component that is a set of pixels connected to each pixel located on the trajectory, Label this connected component. By analyzing the shape of the connected component and the correlation distance in each label, the final recognition of a defect existing on the sealing surface 81a of the sealing member 81 is performed. FIG.
FIG. 4 is a diagram showing an image obtained by examining the eight-point connection of each pixel on the defect trajectory, expanding the defective pixel, and labeling the connected component. As shown in FIG. 14, the defect is labeled in three large areas, and the sealing surface 81a of the sealing member 81
Can be globally specified.

On the other hand, FIG. 16 is a view showing a connected image of each pixel located on the trajectory based on the tracking result of the defect trajectory by the eight-point connection shown in FIG. As can be seen from FIG. 16, when the connected components of each pixel located on the trajectory are obtained based on the tracking result of the defect trajectory by the eight-point connection, there are many connected components,
The shape of the defect can only be captured locally, and it is difficult to accurately recognize the shape of the defect.

As described above, according to the present embodiment, a binarized image is obtained by screening the binarized image data by the lattice filter K based on the threshold T appropriately determined. Therefore, the edge element can be detected with high accuracy by relatively simple processing. In particular, for example, such as an image of a defect existing on the surface of a rubber material, a large amount of sesame salt noise due to the characteristics of the surface of the rubber material is included, and an image having a small difference in shading of the defect image has high accuracy. It is possible to detect a stable edge element.
Further, according to the present embodiment, the trajectory of the defect is traced by searching for a direction in which the number of defective pixels is maximum within a predetermined range from one point of an edge element (a set of defective pixels) of the binary image. By examining the local connectivity of each pixel on the obtained trajectory and extracting defects, defects can be globally extracted, and defect images with a small difference in shading and defects of complex shapes can be extracted. The whole image can be reliably extracted.

[0043]

According to the present invention, for example, as in the case of an image of a defect existing on the surface of a rubber material, the defect image contains a large amount of noise in the image and has a relatively small density difference of the defect. However, a binarized image with good edge element accuracy can be obtained.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration example of a surface inspection apparatus to which an image processing inspection apparatus according to the present invention is applied.

FIG. 2 is a cross-sectional view showing a structure of a seal member 81 for a gas valve as a test object of the surface inspection device 61 shown in FIG.

FIG. 3 is a configuration diagram illustrating a configuration of an image processing inspection device 75;

FIG. 4 is a diagram for explaining processing performed on image data binarized by a filter unit 26;

FIG. 5 is a diagram for explaining the principle of edge tracking of the edge tracking processing unit 31.

FIG. 6 is a diagram for explaining connection of pixels.

FIG. 7 is a flowchart illustrating an example of a processing procedure of the image processing inspection method according to the present invention.

FIG. 8 is a flowchart illustrating an example of a processing procedure following FIG. 7;

9 is a diagram illustrating an example of an image of a seal surface 81a of a seal member 81 input to an image processing inspection device 75. FIG.

FIG. 10 is a diagram illustrating an example of a binarized image obtained by binarizing the image illustrated in FIG. 9;

FIG. 11 is a diagram showing a state where a lattice filter K is set for a binarized image.

FIG. 12 is a diagram showing a binarized image screened by a lattice filter K.

13 is a diagram illustrating a trajectory obtained by a defect tracing process performed on the binarized image illustrated in FIG. 12;

FIG. 14 is a diagram showing an image obtained by examining the eight-point connection of each pixel on the defect trajectory, expanding the defective pixel, and labeling the connected component.

FIG. 15 is a diagram illustrating a trajectory obtained as a result of performing a tracking process of a defect trajectory by examining eight-point connection of each pixel of the binary image illustrated in FIG. 12;

FIG. 16 is a diagram showing a labeled image in which a connected component of each pixel located on the trajectory is obtained based on the tracking result of the defect trajectory by the eight-point connection shown in FIG.

FIG. 17 is a configuration diagram illustrating an example of a surface inspection apparatus.

[Explanation of symbols]

 Reference Signs List 20 binarization processing unit 21 image data input unit 22 density gradient calculation unit 23 average value calculation unit 24 standard deviation calculation unit 25 threshold calculation unit 26 edge element extraction unit 27 filter unit 30 Defect extraction unit 30 31 Edge tracking unit 32 Pixel connection investigation unit 33 Labeling unit

Continued on the front page F term (reference) 2F065 AA12 AA49 BB05 CC00 DD06 FF42 FF66 GG02 GG17 HH02 HH12 HH13 HH14 JJ03 JJ09 JJ26 LL49 PP01 PP13 QQ00 QQ03 QQ07 QQ08 QQ13 QQ25 QQ28 QQ31 Q02 Q41 Q01 Q34 Q01 Q34 A EB01 EC05 ED08 ED22 5B057 AA01 BA02 BA19 BA29 BA30 CA02 CA08 CA12 CA16 CB02 CB06 CB12 CB16 CC03 CE02 CE06 CE12 CH08 DA03 DA08 DA17 DB02 DB05 DB08 DC05 DC14 DC17 DC34 DC36 5L096 AA03 AA07 BA03 FA20 FA04 GA09 GA51 GA55

Claims (2)

[Claims] An image processing inspection method for binarizing image data of a surface of an inspection object and extracting a defect existing on the surface of the inspection object based on the binarized image data. Obtaining a density gradient of each pixel of the image data of the surface of the inspection object; obtaining an average value and a standard deviation of the density gradient of each pixel; and, based on the average value and the standard deviation, A step of determining a threshold value for determining an edge at which the density of the image data changes; a step of binarizing the image data based on the threshold value to extract an edge element including a defect; Screening the quantified image data with a mesh filter having a plurality of square lattices of a predetermined size to remove noise components included in the edge elements. Image processing inspection method for extracting a defect existing on the surface of the inspection object on the basis of the screened image data. 2. An image processing / inspection apparatus for binarizing image data of a surface of an object to be inspected and extracting defects existing on the surface of the object to be inspected based on the binarized image data. A density gradient calculation unit that calculates a density gradient of each pixel of the image data on the surface of the inspection object; an average value calculation unit that calculates an average value of the density gradient of each pixel; and a standard density gradient of each pixel. A standard deviation calculator for determining a deviation, a threshold calculator for determining a threshold for determining an edge where the density of the image data changes based on the average value and the standard deviation, A binarization processing unit for binarizing the image data to extract an edge element including a defect, and screening the binarized image data with a mesh filter having a plurality of square lattices of a predetermined size. And said d Has a filter unit that removes a noise component contained in di-element, the image processing inspection apparatus for extracting a defect existing on the surface of the inspection object on the basis of the screened image data.