JPH10208066A - Method for extracting edge line of check object and appearance check method using this method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an edge line of high precision by specifying an edge point based on an edge trace start candidate point in accordance with peripheral picture elements, their density changes, directions of these changes, etc. SOLUTION: A scan window W1 is scanned to find a picture element having an edge flag point; and if a differential absolute value |e| which this picture element has is larger than a preliminarily set threshold Z1, this point is taken as an edge trace start candidate element Q1. A window for trace of n×m picture elements is set with this edge trace start candidate point Q1 as the center, and a total sum N1 of differential absolute values |e| which edge flag points existing in this window for trace have is obtained; and if this total sum N1 is larger than a preliminarily set threshold Z2, this point is judged to be an edge of a check object, and this edge trace start candidate point is regarded as an edge point. The next edge point is traced based on this edge point Q2.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for extracting an accurate edge line by processing an image of an object to be inspected taken by an image input device by a special method. In particular, the present invention relates to a visual inspection method for detecting a defect such as a chip, a crack, a dirt, or a foreign matter that occurs in an edge portion where it is not easy to identify a corner or the like.

[0002]

2. Description of the Related Art A method of detecting an edge line by using a binary image obtained by photographing the appearance of an inspection object by using a process and detecting a defect of the inspection object based on the edge line is as follows. It is conventionally known. However, in such a method using a binary image, it is difficult to set a threshold value for obtaining a binary image, and it is difficult to obtain an accurate edge line depending on photographing conditions. It was impossible.

[0003] The applicant of the present invention has proposed Japanese Patent Publication No. Hei 7-1881.
No. 1 proposed a defect inspection method based on an edge line obtained by differentiating a multi-tone digital image. The defect inspection method according to this proposal is based on a digital image obtained by converting appearance image information of an inspection object into multiple gradations,
After obtaining an edge image based on the differential absolute value of each pixel in the digital image processing, a search line S is further set, and an edge flag point at which the search line S intersects with the contour line L1 included in the edge image. (Pixel) F, and a plurality of pixels S1, S2, R1, R2 before and after the edge flag point F are extracted as a stick mask, and predetermined arithmetic processing is performed.
This is to detect whether or not there is a defective portion such as a chip or a foreign substance on a search line set inside the inspection object.

FIG. 23 is a flowchart showing a basic procedure in the image processing, and FIGS. 24 (a) and 24 (b) explain this method. In FIG. 24 (a), P4 is an edge image and S is a search line. , L1 indicate the outline of the defect, L0 indicates the outline of the object to be inspected, and FIG.

[0005]

In the appearance inspection method using the stick mask, the density of the defective portion and the density of the peripheral portion, which are provided in the method of determining the defective portion based on the previous binary image, are reduced. The problem that the defect cannot be identified when the defect is included in the same region with respect to the threshold value was solved by analyzing the density change near the defect part, and the defect detection accuracy could be improved.

However, according to this method, the contour of the inspection object is slightly different due to the difference in the mold and the molding conditions, or the contour of the inspection object is sharp, such as a corner portion, even though they are of the same kind. For the portion where the direction changes, the edge flag image itself obtained by the image processing (in the above-described conventional example, the edge image is used) may not accurately reflect the actual contour of the inspection object. As a result, the external shape of the inspection object becomes smaller, and furthermore, if there is a small burr in the external appearance or if the outline of the external appearance is rounded, it is still impossible to detect a stable and highly accurate defect, A more accurate appearance inspection method has been desired.

[0007] The present invention aims to solve the above problems,
The present inventors have arrived as a result of intensive studies, and based on the original image obtained by imaging and processing the inspection object, based on the differential absolute value image, the differential direction value image, and the edge flag image. By finding a new method of searching for edge points by focusing on the change in the density of the image and the direction of the change, the new edge line of the inspection object that can obtain a more accurate edge line Based on the extraction method and the edge line obtained by the method, various inspection lines, inspection areas, arithmetic processing, etc.
By performing digital image analysis, it is possible to provide a highly accurate and stable appearance inspection method.

[0008]

According to the present invention, in order to achieve the above object, an original image obtained by imaging and processing an object to be inspected, and a differential absolute value obtained by preprocessing the original image. Based on the image or the differential direction value image, and the edge flag image obtained by the thinning processing and the edge flag extension processing based on the differential absolute value image, the edge line can be formed by using a further new method. Extract.

In the first method of extracting a new edge line (claim 1), an edge flag point selected in an edge flag image is set as an edge tracking start candidate point, and the edge absolute value of the surrounding edge flag points is calculated by summing the differential absolute values of the surrounding edge flag points. The edge tracking start candidate point is regarded as an edge point, and a pixel adjacent to the edge point and having a differential absolute value exceeding a predetermined threshold value is set as a next edge point, and the edge point is set as a reference. The edge line is extracted by repeating the same processing.

A second method of extracting a new edge line (claim 2) is to use an edge flag point selected in the edge flag image based on a differential absolute value and a differential direction value as an edge point, and with the edge point as a reference, An adjacent pixel whose absolute differential value exceeds a predetermined threshold value and whose differential direction value is within a predetermined range is set as the next edge point, and the same processing is repeated with the edge point as a reference to repeat the edge line. Is extracted.

Based on the edge line thus obtained, an inspection line having the same shape as the edge line, a plurality of equally spaced inspection lines in one direction or a direction perpendicular to the inside of the edge line are used. Alternatively, the density of the surrounding pixels, the average density, the differential absolute value, the differential direction value, or the sum of them is used as appropriate. If the extracted edge lines are two edge lines that are close to each other at a fixed interval, An inspection area is appropriately set between the two edge lines to detect a defect of the inspection object.

A method for inspecting the appearance of an object to be inspected, which is proposed at the same time as the present invention, comprises: setting an inspection line on the basis of an edge line extracted using the first and second edge line extraction methods described above; Various methods are used to detect defects of the inspection object. Claims 3 to 6 are first and second.
The method is applied to the case where the number of edge lines extracted by the edge line extraction method is one, and the method according to claim 3, wherein a pixel on an inspection line set based on the extracted edge line or a pixel adjacent thereto is determined by a known method. And the presence or absence of a defect is determined. In claim 4, an inspection line having the same shape is set for the extracted edge line, and neighboring pixels on the inspection line are analyzed by a known method.
The presence or absence of a defect is determined. Further, in the fifth aspect, the defect of the inspection object is determined based on the density difference of the pixel and the differential absolute value.

On the other hand, according to claims 7 to 10, the edge lines extracted by the first and second edge line extraction methods are two or more.
This method is applied to a case where a book is used.
At least one pixel existing between the two edge lines is to be analyzed in determining whether or not there is a defect. In claim 8, all the pixels in the region included in the two edge lines are analyzed. In claim 9,
A midpoint inspection line is set between two edge lines, and pixels on the midpoint inspection line are to be analyzed. In the tenth aspect, a plurality of inspection lines are set between two edge lines. Each of the methods proposes a method of setting and analyzing pixels on the inspection line.

Further, claim 11 to claim 15 are the first,
Based on the edge line extracted by the second edge line extraction method, as a result of the inventor's earnest study,
The proposed defect detection method is proposed. That is, Claims 11 and 12 describe a defect on the edge line of the inspection object,
A method for simultaneously detecting external defects beyond the edge line, claims 13 and 14 are appearance inspection methods using phase shift, and claims 15 are based on pixels on the inspection line.
A method of setting a window and comparing the average densities of the windows with each other to detect a defect has been proposed.

According to the present inventor, as a result of performing the appearance inspection of the object to be inspected by the method proposed above, the detection accuracy of the defect of the object to be inspected is remarkably improved to 3 to 5 times as compared with the conventional method. They have found out and have proposed the method of the present invention.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION

[Embodiment 1] In Embodiment 1, a first edge line extraction method will be described. In this method, as shown in FIG. 3, an object to be inspected is imaged by an image input device 1 such as a television camera, and the density of each pixel is converted into a digital signal by an analog-to-digital converter 2. By performing the following preprocessing in, a differential absolute value image, a differential direction value image, and an edge flag image are obtained in addition to the original image obtained by the analog-digital converter 2.

That is, the original image P0 obtained by imaging the spatial region including the inspection object is a grayscale image,
As shown in (a), the inspection object O, defect X1, foreign matter X2
It is an image including such. The process of extracting the edge flag such as the contour line of the inspection object O from the grayscale image is as follows.
It is based on the idea that "the edge corresponds to a portion where the density change is large". Therefore, the edge flag is extracted by differentiating the density. The differentiation process is performed by scanning the original image P0 with a local parallel window W of 3 × 3 pixels as shown in FIG. That is, a pixel E of interest and eight pixels (near eight) A around the pixel E
To D and F to I, a local parallel window W is formed, and a vertical density change ΔV and a horizontal density change ΔH of the density of the pixels A to I in the local parallel window W are obtained by the following equation. = (Da + Db + Dc)-(Dg + Dh + Di) ΔH = (Da + Dd + Dg)-(Dc + Df + Di) Further, the differential absolute value | e | and the differential direction value ∠e are obtained by the following equations.

[0018]

(Equation 1)

[0019]

(Equation 2)

Here, Da to Di indicate the density of the corresponding pixel. In addition, when the density is referred to here, it indicates the density of white, and the higher the density, the brighter it is. When the direction of the density change, here the direction of the change from white to black, is determined by a hexadecimal value as shown in FIG. 6A, the value of the differential direction is “4” (90 degrees in angle). ) Is added. For example, in the case of the example shown in FIG. 6B, the change direction of the density in this figure is “0”, and when this figure is processed, an edge flag Q as shown in FIG. 6C is obtained. The differential direction value of the edge flag Q is “4”. In the case of FIG. 6D, the change direction of the density is “4” and the differential direction value is “8”.

As is apparent from the equations (1) and (2), the differential absolute value | e | represents the maximum change rate of the density in the vicinity of the pixel of interest in the original image, and the differential direction value {e
Represents a direction orthogonal to the direction of the maximum change in density in the vicinity region. By performing the above operation on all the pixels of the original image P0, a portion having a large density change such as the contour of the inspection object O, the defect X1, the foreign matter X2, etc., and the direction of the change are extracted. be able to.
Here, an image representing the differential absolute value | e | of each pixel is referred to as a differential absolute value image P1, and an image representing the differential direction value ∠e is referred to as a differential direction value image P2.

Next, a thinning process is performed. The thinning process is performed while paying attention to the fact that the larger the absolute differential value, the larger the density change. That is, the differential absolute value of each pixel is compared with the differential absolute values of the surrounding pixels, and those that are larger than the surrounding pixels are connected, whereby 1
An edge flag having a pixel width is extracted. As shown in FIG. 7, considering the differential absolute value three-dimensional schematic image P5 in which the position of each pixel is represented by XY coordinates and the differential absolute value is taken on the Z axis, the thinning processing calculates the ridge line on this curved surface. It is equivalent to By the processing so far, all edges are extracted regardless of the magnitude of the differential absolute value. Since the ridgeline obtained at this stage includes unnecessary small peaks due to noise or the like, a threshold value ZL is appropriately set as shown in FIG. To remove noise components. The image obtained by this process is
When the contrast of the original image is insufficient or when there is much noise, discontinuous lines are likely to occur. Therefore,
Perform edge flag extension processing.

In this edge flag extension processing, starting from the end point of the discontinuous line, the pixel of interest is compared with the surrounding pixels, and the line is drawn in the direction in which the evaluation function f (ej) expressed by the following equation is the largest. And continue this until it hits another line endpoint.

[0024]

(Equation 3)

Here, ∠e0 is the differential direction value of the center pixel (corresponding to E of the local parallel window W), and | ej |
, ∠ej (j = 1,2,3,4,5,6,7,8) are differential absolute values and differential direction values of adjacent pixels (corresponding to 8 neighborhoods of the local parallel window W). By the above processing, as shown in FIG. 4B, an edge flag image P3 which traces a portion where the density change is large in the original image is obtained. In the edge flag image P3, each line is the inspection object O, the defect X1, the foreign matter X2.
Can be regarded as representing the edge flag lines L0 to L2 corresponding to the contour lines.

By the above preprocessing, four types of images are obtained: an original image P0, a differential absolute value image P1 (not shown), a differential direction value image P2 (not shown), and an edge flag image P3.
Each image is stored in an original image memory 4, a differential absolute value image memory 5, a differential direction value image memory 6, and an edge flag image memory 7, respectively. The original image is a gray-scale image, and the density is usually represented by 8 bits.
Is 0 ≦ a ≦ 255. The value b of each pixel of the differential absolute value image (that is, the differential absolute value) is represented by, for example, 6 bits, and 0 ≦ b ≦ 63, and the value of each pixel of the differential direction value image (that is, the differential direction value) Since c) represents 16 directions by hexadecimal notation, 0 ≦ c ≦ 15. For the edge flag image, it is only the presence or absence of a line,
Pixels serving as lines are represented as “1”, and other pixels are represented as “0”. That is, the value range is {0, 1}.

In the present invention, based on the edge flag image generated by the above procedure, digital image analysis is performed by the following procedure to extract an edge line. FIG. 1 is a flowchart showing the flow of the process of the first method. The pre-processing has already been described, and will not be described. First, a scanning window W1 (see FIG. 9) is set arbitrarily in the edge flag image P3. Next, the scanning window W1 is scanned to find a pixel having an edge flag point, and the differential absolute value of the pixel is found.
If | e | is greater than a preset threshold Z1,
This point is defined as an edge tracking start candidate point Q1.

Next, a tracking window W2 of n × m pixels is set around the edge tracking start candidate point Q1 (see FIG. 10), and the differential absolute value of the edge flag point existing in the tracking window W2 is set. The sum N1 of | e | is determined. If the sum N1 is larger than a predetermined threshold value Z2, the point is determined to be the edge of the inspection object, and this edge tracking start candidate point Q1 is determined as an edge point. Regarded as Q2.

Next, based on the edge point Q2,
The next edge point is tracked (see FIG. 11). Edge point Q
A tracking window W2 is set centered at 2, and a search is performed to find a pixel adjacent to the edge point Q2 and having a differential absolute value | e | greater than a predetermined threshold value Z1. When a pixel that satisfies the above condition is found, that point is set as the next reference edge point, the tracking window W2 is set again around the reference edge point, and the next edge point is tracked in the same manner. . In this tracking, the obtained edge point goes out of a predetermined inspection area (not shown), or
The process is terminated when a pixel meeting the above condition cannot be found in the tracking window W2.

In the process of tracking the edge points, if there is no pixel that satisfies the above condition among the pixels adjacent to the reference edge point, edge connection processing is performed. The edge point Q3 shown in FIG. 11 is such an edge point. In this case, a search is performed for a pixel that is one pixel away from the edge point Q3.
Next, a search is made to see if there is any pixel that satisfies the above condition, while expanding the range in order of pixels three pixels away. This search is continued until the pixel at the search point goes out of the tracking window W2. As a result, when a pixel that satisfies the above condition is found, that point is set as the next reference edge point Q4, and the connection is made between the edge points Q3 and Q4 assuming that there is an edge therebetween. If no pixel satisfying the above condition is found, the tracking ends.

The edge points obtained by such tracking are sequentially connected and extracted as an edge line of the inspection object. In the above method, the edge point is identified while sequentially determining the threshold value of the adjacent pixel. However, it is not always necessary to preferentially identify the adjacent pixel as the edge point, and the edge point is identified during the identification. In, a pixel that is easy to specify may be specified as an edge point first.

[Second Embodiment] In a second embodiment, a second method of extracting an edge line (corresponding to the method described in claim 2) will be described. FIG. 2 is a flowchart showing the flow of the process of the second method. Among them, the pre-processing has already been described in the first embodiment. In the second method, digital image analysis is performed by the following procedure to extract an edge line.

First, in the edge flag image P3, a scanning window W1 (see FIG. 9) is arbitrarily set. The scanning window W1 is scanned to find the edge flag point, and when the differential absolute value | e | of the pixel is larger than a predetermined threshold value Z1, and the differential direction value ∠ When e is within a preset range, that point is set as an edge point Q2.

Next, based on the edge point Q2,
The next edge point is tracked (see FIG. 11). That is,
A tracking window W2 is set around the edge point Q2, and a predetermined threshold value Z adjacent to the edge point Q2 is set.
A search is performed to find a pixel that has a differential absolute value | e | greater than 1 and that has a differential direction value ∠e within a preset range. When a pixel that satisfies the above condition is found, that point is set as the next reference edge point, a tracking window W2 is set around the reference edge point, and the next edge point is tracked in the same manner. I do. This tracking
The process ends when the obtained edge point goes out of the inspection area (not shown) set in advance or when a pixel satisfying the above condition cannot be searched in the tracking window W2.

In the process of tracking the edge points, if there is no pixel that satisfies the above condition among the pixels adjacent to the reference edge point, edge connection processing is performed. The edge point Q3 shown in FIG. 11 is such an edge point. In this case, a search is performed for a pixel that is one pixel away from the edge point Q3.
Next, a search is made to see if there is any pixel that satisfies the above condition, while expanding the range in order of pixels three pixels away. This search is continued until the pixel at the search point goes out of the tracking window W2. As a result, when a pixel that satisfies the above condition is found, that point is set as the next reference edge point Q4, and the connection is made between the edge points Q3 and Q4 assuming that there is an edge therebetween. If no pixel satisfying the above condition is found, the tracking ends.

Edge points obtained by such tracking are sequentially connected and extracted as an edge line of the inspection object. Third Embodiment Next, with reference to the edge line of the inspection object extracted by the first and second edge line extraction methods, an inspection line having the same shape as the edge line or one direction inside the edge line or Using a plurality of equally spaced inspection lines in the orthogonal direction, the density, average density, differential absolute value,
An appearance inspection method for detecting a defect of an inspection object by appropriately using the differential direction values or the sum of the values will be described below.

In the first appearance inspection method (corresponding to the third and fourth aspects), first, the edge lines extracted by the first and second edge line extraction methods represent the contours of the inspection object. Therefore, the densities of the pixels inside and outside the edge line are compared, and the side with the higher density is determined as the outside of the test object, and the side with the lower density is determined as the inside of the test object. Next, as shown in FIGS. 12 and 13, while tracing the extracted edge lines R1 and R2, the inspection lines S1 and S2 are set at locations separated by a predetermined pixel d inside the inspection object. Go. A plurality of inspection lines S1 and S2 can be set at equal intervals.

The inspection lines S1 and S2 may be curved as shown in FIG. 12 or straight as shown in FIG. 13, and may be of the same type according to the inspection object having an arbitrary contour. However, even if the contour of the inspection object is slightly different due to differences in the mold and molding conditions, or even if the inspection object has a contour line that changes direction rapidly, such as a corner portion, Edge lines R1 and R2 that faithfully reflect the outline
It can be set according to.

Thus, the edge line R of the inspection object is
1, inspection lines S1 and S2 having the same shape as R2 are set, and the presence or absence of a defect is determined by a known method for analyzing pixels on the inspection lines S1 and S2. In the case where the inspection lines S1 and S2 are used as the basis, a method using a density value measurement window W3 as shown in FIG. 14 (corresponding to the method described in claims 3 and 5) is particularly used as a defect detection method. explain.

As shown in FIG. 15, the inspection lines S1, S2
By scanning the density value measurement window W3 above, the central pixel C0 in the density value measurement window W3, and the density measurement pixels A1, A2, A density difference between B1 and B2 is obtained and compared with a predetermined threshold value Z3. That is, the pixel A for density value measurement
The concentrations of 1, A2, B1, and B2 are Da1, Da2, Db1, and D2.
Assuming that b2, when the equation | Da1-Db1 | + | Da2-Db2 |> Z3 holds, "1" is added to the count K with the center pixel C0 as a defect candidate point in the defect candidate portion X0. This process is performed while scanning the inspection lines S1 and S2. When the count K is finally larger than the predetermined threshold value Z4, it is determined that there is a defect.

In the above-described method using the density measurement pixels, in addition to the density difference, the density measurement pixels A1, A2,
When the sum N2 of the differential absolute values | e | of B1 and B2 is larger than a predetermined threshold value Z5, a method of determining that there is a defect (corresponding to the method of claim 3 or 6) is also defect detection. It is effective for [Embodiment 4] Next, a visual inspection method effective when the edge lines extracted by the first edge line extraction method are extracted as two edge lines that are close to each other at a constant interval. 9 and 10).

As shown in FIG. 16 (a), when the density is particularly high near the contour of the object O, that is, when the edge portion Q5 has a higher luminance than the density of the background Y and the object O, As shown in FIG. 16B, the edge lines extracted by the method of the first embodiment may be extracted as two edge lines RW adjacent to each other at a certain interval. In this case, considering a case where a defect occurs between the edge lines RW, the inspection window W as shown in FIG.
4, and a defect X5 is detected by setting a region sandwiched between the edge lines RW as an inspection region W5.
8).

In the method of the third embodiment, in such a case, the inspection line cannot be set clearly between the edge lines, and the method proposed here is particularly effective. When the interval between the edge lines RW is wide, as shown in FIG. 17, a midpoint inspection line S3 is set at the center of the edge line RW, and the inspection line S3 is set as shown in FIG. Example 3 using the indicated density value measurement window W3
A method for detecting the defect X6 in the same manner as described above.
9).

If the interval between the edge lines RW is even wider, a plurality of inspection lines S4 are set between the edge lines RW, as shown in FIG.
The method of detecting the defect X7 in the same manner as in the third embodiment using the density value measurement window W3 shown in FIG. 14 of the third embodiment (corresponding to the method described in claims 7 and 10). Be taken.

[Embodiment 5] Based on the edge lines extracted by the first and second edge line extraction methods,
An appearance inspection method for setting an inspection line and detecting a defect occurring inside the inspection object and around the edge line will be described. In this method (corresponding to the method described in claim 11), as shown in FIG. 19A, in a predetermined inspection area W6 including the extracted edge line R3, one direction is provided inside the edge line R3. Inspection line S5 at equal intervals
Set multiple lines.

Next, in this inspection area W6, the inspection line S5 is scanned and a defect is detected by the defect detection method using the density value measurement window W3 used in the third embodiment. In this case, at the point where the inspection line S5 and the edge line R3 intersect, FIG.
When there is no defect near the edge line R3 as in
Although the inspection line S5 ends at the intersection, FIG.
(B) When the defect X8 is present near the edge line R3 as shown in FIG. 7B, the differential absolute value | e | | e | or the differential direction value ∠e. Therefore, if the differential absolute value | e | at the intersection is smaller than a predetermined threshold value or the differential direction value ∠e is not in a predetermined range,
As shown in FIG. 19C, the inspection line S5 is extended beyond the edge line R3 to another side of the inspection area W6, and inspection is performed on the inspection line S5.

This method is suitable for detecting defects inside the inspection object and around the edge line. Further, depending on the shape of the edge line of the inspection object, FIG.
As shown in (b) and (c), a method of setting the inspection line S5 not only in one direction but also in two orthogonal directions (corresponding to the method according to claim 12) is adopted. In this figure, W7, R4, S6, and X9 indicate the inspection area, edge line, inspection line, and defect in this case.

[Embodiment 6] Defects occurring in an inspection object are detected by shifting the phase of an inspection line based on the edge lines extracted by the edge line extraction methods of the first and second methods. The appearance inspection method will be described. This method (corresponding to the method according to claim 13) sets an inspection line having the same shape as the edge line as in the third embodiment, but is characterized in that the inspection line is shifted in phase.

First, as shown in FIG. 21A, an inspection line S7 is set inside the edge line R5 at an arbitrary pixel interval. Next, the phase of the inspection line S7 is shifted. The phase shift refers to moving the inspection line S7 by a predetermined number of pixels before or after the direction of the target edge line R5.
For example, as shown in FIG.
Is circular-shaped, it is rotated about the center of curvature of the circular arc. When the edge line is a straight line, it moves in parallel. Thereafter, by scanning the inspection line after the movement, a defect is detected.

This method is not effective when there is no defect on the edge line R5 as shown in FIG.
This is adopted when there is a defect X10 on the edge line R5a as shown in FIG. In this case, do not shift the phase,
It is difficult to detect a defect simply by scanning on the inspection line S7a, since a change in density indicating a defect, a characteristic of a differential absolute value | e | or a differential direction value 得 e cannot be obtained on the line. It is. Therefore, when the phase is shifted as shown in FIG. 21C, edge points Q6 and Q7, which are the intersections of the edge line R5a and the inspection line S7b after the movement, are obtained. As is clear from FIG. Since there is a density difference before and after the edge points Q6 and Q7 due to the defect X10, the inspection line S7a after the movement is scanned, and the defect is detected by the method using the density value measurement window W3 used in the third embodiment. Can be detected.

As described above, by performing the phase shift, it is possible to detect a minute defect occurring around the edge line. Further, in the above method, an inspection line is provided not only inside the edge line but also outside the edge line, and the phase is shifted by the same method to detect a defect (claim 14). Method)
It is effective when there is a defect outside the edge line, for example, when there is a burr or the like.

[Embodiment 7] Finally, a defect is detected by scanning the average density measurement window on the inspection line set based on the edge lines extracted by the first and second edge line extraction methods. The appearance inspection method (corresponding to the method according to claim 15) will be described. This method is an appearance inspection method in which an inspection line having the same shape as an edge line is set as in the third embodiment, and a defect is detected by scanning the inspection line. An average density measurement window is used.

As shown in FIG. 22A, an average density measurement window W9 of n × m pixels is used, and as shown in FIG. 22B, the average density is measured on the inspection line S8 having the same shape as the edge line R6. While sequentially moving the measurement window W9,
With respect to the pixels on the inspection line S8, the average density DW, which is the average value of the densities of the pixels in the average density measurement window W9 centered on the inspection line S8, is sequentially obtained. If the actual difference DS is larger than a predetermined threshold value Z6, it is determined that there is a defect.

According to this method, when the average density measurement window W9 is located at a pixel including the defect X11, the defect X1
In the part 1, the average density DW is increased by including a white part having a high density outside the inspection object, and the relative difference DS of the average density is increased compared to the part without defects. , Defect detection. According to this method, it is possible to detect information not only on the inspection line but also on a defect candidate portion within a range having a certain extent. The method of extracting an edge line based on an original image, a differential absolute value image, a differential direction value image, and an edge flag image of the inspection object, which is proposed by the present invention, is described in the first embodiment. To specify Q2, sum N of differential absolute values | e |
In addition to the method using 1, the average value of the concentration and the differential absolute value | e |
It is also possible to use a method using the average value of the above.

As described in the second embodiment, the method of using the sum N1 of the differential absolute values | e | and the differential direction value ∠e to specify the edge point Q2, other than the method of using the average value of the density and the differential absolute value |
A method using an average value of e | is also possible. Further, other defect detection methods that appropriately combine various inspection lines, inspection areas, differential absolute values, differential direction values, densities, and sums and averages of these, which are proposed in the present invention, are also possible.

[0056]

According to the first and second edge line extraction methods (claims 1 and 2), an original image, a differential absolute value image, a differential direction value image, and an edge flag image of an inspected object are used. Then, from the pixels of the edge flag image, an edge tracking start candidate point is specified, and based on the edge tracking start candidate point, an edge point is specified from a density change with respect to surrounding pixels or a direction of change. Therefore, it is possible to improve the accuracy even for the inspection object that has the same type, but the contour is slightly different due to the difference in the mold and the molding conditions, and the part where the direction changes rapidly such as the corner. An edge line can be obtained.

Further, according to the appearance inspection method (claims 3 to 15) of the present invention proposed at the same time, the first and second appearance inspection methods described above are provided.
Based on the accurate edge line extracted by the edge line extraction method, the inspection line set based on the edge line or the pixels near the inspection line are analyzed to determine whether there is a defect. Therefore, it is possible to detect minute defects such as peeling chips generated inside without inspecting the shape of the peripheral portion of the edge line of the inspection object.

In particular, according to the appearance inspection method proposed in claims 3 to 6, an inspection line is specified based on the edge lines extracted by the first and second edge line extraction methods, and a known method is used. Is used to detect a defect, so that the defect can be detected easily with high accuracy. Also, according to the appearance inspection method proposed in claims 7 to 10, the edge lines extracted by the first and second edge line extraction methods are mutually spaced at a certain interval because the edge portions are high in brightness and thick. Even when two adjacent edge lines are extracted, stable inspection can be performed.

According to the appearance inspection method proposed in claims 11 and 12, the inspection set inside the inspection object based on the edge lines extracted by the first and second edge line extraction methods. If there is a defect near the edge line, the line is extended beyond the edge line to the outside of the object to be inspected, so that the inside and outside of the edge line of the object (inside and outside of the object) are inspected. ) Can also be detected with high accuracy.

According to the appearance inspection method proposed in claims 13 and 14, the same shape as the edge line extracted by the first and second edge line extraction methods is set inside or outside the edge line. Since the inspection is performed by shifting the phase of the inspection line, it is possible to detect minute defects such as peeling chips generated on and around the edge line.

Further, according to the appearance inspection method proposed in claim 15, the inspection line set inside the inspection object based on the edge lines extracted by the first and second edge line extraction methods. Is inspected while scanning the density value measurement window having a predetermined range, so that not only the inspection line but also the peripheral defects can be detected.

Further, according to the present invention, the shape of the inspection object,
Since an appropriate method can be selected from the proposed inspection methods according to the characteristics of the edge line, the accuracy of defect detection as a whole is remarkably improved to 3 to 5 times as compared with the conventional method.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a processing procedure in an edge line extraction method (claim 1) of the present invention.

FIG. 2 is a flowchart illustrating a processing procedure in the edge line extraction method (claim 2) of the present invention.

FIG. 3 is a block diagram of a processing circuit for implementing the method of the present invention.

FIGS. 4A and 4B are explanatory diagrams showing examples of an original image and an edge flag image, respectively.

FIG. 5 is a conceptual explanatory diagram of a local parallel window used in differential processing.

6A is a conceptual explanatory diagram of a differential direction value, FIG. 6B is an explanatory diagram showing an example of a density change, FIG. 6C is an explanatory diagram after image processing in a similar case, and FIG. Illustration after image processing in case

FIG. 7 is a schematic diagram showing a differential absolute value three-dimensionally.

FIG. 8 is a conceptual explanatory diagram of a noise removal process for obtaining a differential absolute value image.

FIG. 9 is an explanatory diagram of an edge tracking start candidate point used in the method of the present invention.

FIG. 10 is an explanatory diagram of a tracking window.

FIG. 11 is an explanatory diagram of edge connection processing.

FIG. 12 is a diagram showing an example of an inspection line (circular shape) set for an edge line extracted by the method of the present invention.

FIG. 13 is a diagram showing an example of an inspection line (linear shape) set for an edge line extracted by the method of the present invention.

FIG. 14 is an explanatory diagram of a density value measurement window used to determine a defect.

FIG. 15 is an explanatory diagram showing a relationship between a density value measurement window and a defect.

FIGS. 16A to 16C are diagrams for explaining processing procedures (claims 7 and 8) of the method of the present invention when two edge lines that are close to each other at a fixed interval are extracted;

FIG. 17 is a diagram for explaining an inspection line (claims 7 and 9) of the method of the present invention when two edge lines that are close to each other at a fixed interval are extracted.

FIG. 18 is a view for explaining an inspection line (claims 7 and 10) of the method of the present invention when two edge lines which are close to each other at a fixed interval are extracted.

19 (a) to 19 (c) are views for explaining an appearance inspection method according to the present invention proposed in claim 11. FIG.

FIGS. 20A to 20C are diagrams for explaining the appearance inspection method of the present invention proposed in claim 12;

FIGS. 21A to 21C are diagrams for explaining an appearance inspection method (an appearance inspection method using a phase shift) of the present invention proposed in claims 13 and 14;

FIGS. 22A and 22B are diagrams illustrating an inspection window and an inspection line used in the appearance inspection method of the present invention proposed in claim 15;

FIG. 23 is a flowchart illustrating a basic processing procedure of an appearance inspection method using a stick mask (Japanese Patent Publication No. 7-18811).

24 (a) and (b) are explanatory diagrams of a basic concept of an appearance inspection method using a stick mask (Japanese Patent Publication No. 7-18811).

[Explanation of symbols]

 P0: Original image P3: Edge flag image P5: Differential absolute value stereoscopic image X, X0 to X11: Dust, defect candidate part, defect, foreign matter L0 to L2: Outline line O. ..Inspection object W, W1 to W9: window or inspection area Q: edge flag R: edge line S: inspection line Y: background

Claims (15)

[Claims] An image of an object to be inspected is obtained based on an original image obtained by imaging and processing an object to be inspected and a differential absolute value image obtained by processing each pixel included in the original image. A method for extracting an edge line, comprising: generating an edge flag obtained by performing a thinning process based on a differential absolute value image as an edge flag image; Is set, and within the scanning window, an edge flag point whose differential absolute value exceeds a predetermined threshold value is specified as an edge tracking start candidate point. Then, a predetermined tracking is performed around this edge tracking start candidate point. A window for calculating the sum of the differential absolute values of the edge flag points included in the window.
When the sum exceeds a predetermined threshold, the edge tracking start candidate point is specified as an edge point. After the edge point is specified in this manner, the edge point is determined based on the edge point. Of the pixels adjacent to
A pixel whose differential absolute value exceeds a predetermined threshold value is specified as the next edge point, and further, based on the edge point,
An edge line extraction method for an object to be inspected, wherein an edge line is extracted by sequentially and repeatedly performing a process of specifying a next edge point. 2. An original image obtained by imaging and processing an object to be inspected, and a differential absolute value image and a differential direction value image obtained by processing each pixel included in the original image. A method for extracting an edge line of the inspection object, wherein an edge flag obtained by performing a thinning process based on the differential absolute value image is generated as an edge flag image, and , A predetermined scanning window is set, and an edge flag point whose differential absolute value exceeds a predetermined threshold value and whose differential direction value is within a predetermined range is specified as an edge point in the scanning window. After the edge point is specified, the pixel adjacent to the edge point is determined based on the edge point.
A pixel whose differential absolute value exceeds a predetermined threshold value and whose differential direction value is within a predetermined setting range is specified as the next edge point, and further, the next edge point is determined based on the edge point. An edge line extraction method for an object to be inspected, wherein an edge line is extracted by sequentially and repeatedly performing a specifying process. 3. An object to be inspected by setting a predetermined inspection line based on the edge line extracted in claim 1 or 2, and analyzing neighboring pixels on the inspection line or adjacent pixels including the inspection line. A method for inspecting the appearance of an object to be inspected, comprising detecting a defect of the object. 4. An inspection line having the same shape as at least a part of the edge line is set inside the inspection object identified based on the edge line extracted in claim 1 or 2, and the inspection is performed. A method for inspecting an appearance of an object to be inspected, wherein a defect on the object to be inspected is detected by analyzing pixels on a line. 5. The inspection object according to claim 4, wherein a defect of the inspection object is detected on the inspection line based on a difference in density of pixels adjacent to the inspection pixel. Appearance inspection method. 6. The method according to claim 4, wherein, based on a difference between densities of pixels adjacent to the inspection pixel on the inspection line and a sum of differential absolute values,
A method for inspecting an appearance of an inspection object, comprising detecting a defect of the inspection object. 7. In a case where the edge lines of the object to be inspected extracted in claim 1 are extracted as two edge lines that are close to each other at a fixed interval, a pixel between the two edge lines is analyzed. A defect of the inspection object is detected by the inspection method. 8. The inspection apparatus according to claim 7, wherein a defect existing in the inspection object is detected by analyzing a pixel in the inspection area using an area existing between the two edge lines as an inspection area. Inspection method for the object to be inspected. 9. The defect of an object to be inspected according to claim 7, wherein a midpoint inspection line is set between the two edge lines, and pixels on the midpoint inspection line are analyzed. A method for inspecting the appearance of an object to be inspected. 10. A defect of an object to be inspected according to claim 7, wherein a plurality of inspection lines are set between the two edge lines, and pixels on the plurality of inspection lines are analyzed. A method for inspecting the appearance of an object to be inspected. 11. The method according to claim 1, wherein a predetermined inspection area is set.
Or, by setting a plurality of inspection lines in the same direction having a predetermined interval in the inspection object identified based on the edge line extracted in 2, and analyzing pixels on the inspection line When detecting a defect, at a pixel where the inspection line intersects with the edge line, when the differential absolute value of the pixel is smaller than a predetermined threshold value or the differential direction value of the pixel is not in a predetermined range, An inspection object which extends the inspection line beyond the edge line to one side of the inspection area, and detects a defect of the inspection object by analyzing pixels on the extended inspection line. Appearance inspection method. 12. The method according to claim 1, wherein a predetermined inspection area is set.
Or a plurality of inspection lines having a predetermined interval in a certain direction and a direction orthogonal to the direction are set in a grid inside the inspection object identified based on the edge line extracted in or 2, When a defect is detected by analyzing a pixel on an inspection line, at a pixel where the inspection line and the edge line intersect, the absolute value of the differential of the pixel is smaller than a predetermined threshold value or the differential of the pixel. When the direction value is not within the predetermined range, the inspection line is extended beyond the edge line to one side of the inspection area, and a pixel on the extended inspection line is analyzed to determine a defect of the inspection object. A method for inspecting the appearance of an object to be inspected, characterized by detecting. 13. An inspection object identified on the basis of the edge line extracted in claim 1 or 2,
After setting an inspection line having the same shape as the edge line, shifting the phase of the inspection line, and analyzing pixels on the inspection line shifted in phase, a defect of the inspection object is detected. Inspection method for the object to be inspected. 14. An inspection line having the same shape as the edge line is set inside and outside the inspection object identified based on the edge line extracted in claim 1 or 2, and the inspection line is set. A visual inspection method for an object to be inspected, comprising: detecting a defect of the object to be inspected by analyzing pixels on the inspection line shifted in phase after the phase is shifted. 15. An inspection line having the same shape as the edge line is set inside the inspection object identified based on the edge line extracted in claim 1 or 2, and pixels on the inspection line are set. The method is characterized in that the average density of the pixels in a predetermined window centered is sequentially obtained, and the average density of the pixels separated by a predetermined pixel on the inspection line is compared to detect a defect of the inspection object. Inspection method of the inspected object to be inspected.