JPH06215109A - Picture recognizing method - Google Patents

Abstract

PURPOSE:To extract a smooth circular linear picture for reflecting an original contour line by removing the unnecessary projecting line parts included in the circular linear picture including noises. CONSTITUTION:Circular linear graphic data including projecting curve parts are inputted to a coordinate center deciding processing 11. The processing 11 calculates the intersection of the diagonals of a circumscribed rectangle of the linear graphic form applied as a coordinate system. A polar coordinate system conversion process 12 converts the coordinate system into a polar coordinate system around the intersection as a center. A maximum distance point extracting processing 13 and a smooth section extracting processing 14 extract curve parts equivalent to the circumference part of the circular shape in the line diagram. Then, a projecting part discriminating processing 15 detects the projecting curve parts, and a projecting part compensating processing 16 removes the curved parts, and compensates a part between the curved parts. Thus, the circular linear diagram which does not include any projecting curve part can be obtained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a rivet defect detection system for inspecting a rivet defect in an aircraft rivet using a computer, and removes circular line-shaped protruding noises such as caulking heads of rivets. The present invention relates to an image recognition method for performing accurate recognition.

[0002]

2. Description of the Related Art Conventionally, as a technique in such a field,
For example, some documents were described in the following documents.

Reference: Proceedings of the 42nd National Convention of Information Processing Society of Japan, (Head 3-2-25) Yoshitaka Nishiyama et al. 2-55 to 2-56 As described in the above-mentioned document, the driving quality of rivets, for example, at the time of manufacturing an aircraft plays an important role in the safety of the aircraft. Therefore, an automatic rivet appearance inspection system is being developed that automatically performs an appearance inspection after the rivets are driven. In this rivet appearance inspection system, in the rivet defect detection method, the circular caulking head of the rivet is photographed from directly above with a CCD camera to extract the circular contour area of the rivet. Due to the cause of color unevenness on the upper surface of the rivet, dirt on the plate material (not a defect), and the nature of the LOG filter, the extracted contour area is
Dark areas (projections) other than the original contour area are also included. Since these projecting areas hinder the measurement of the outer diameter of the rivet and the detection of cracks, the contour areas of the rivet should be extracted by removing the projecting areas by the image recognition method. There is. Then, using the contour line of the rivet extracted in this way, the outer diameter of the rivet is measured and cracks are detected.

An example of a conventional image recognition method for a circular object used in such a rivet defect detection method will be described below with reference to FIGS. 6 and 7.

FIG. 6 is a diagram showing processing contents in a conventional image recognition method for a circular object. In the conventional image recognition method, in image input processing 1, grayscale image data is input as an original image from an image input device such as a CCD camera and sent to area extraction processing 2. In the area extraction processing 2, the shape of the target object is extracted as an area on the image from the original image, and the extraction result is sent to the contour line extraction processing 3. In the contour line extraction processing 3, boundary line tracing is performed on the target object area, the contour line of the target object is extracted, and the extraction result is sent to the contour line recognition processing 4. In the contour recognition processing 4, recognition processing such as recognition of a shape result is performed on the obtained contour of the target object.

As described above, in the conventional image recognition method, when the shape of a circular object is recognized, the area extraction processing 2 performs the image processing on the original image to determine the shape of the target object as an area on the image. To extract. Then, in the contour line recognition processing 4, recognition is performed based on the shape of the circular line image obtained by tracing the boundary line of this area in the contour line extraction processing 3. These processes 2 to 4 are executed by a program process using a computer.

In the case of this image recognition processing, the area other than the target object may be connected to the area extracted in the area extraction processing 2 due to the influence of noise or the like. When such a connection occurs, information about the contour line shape of the target object is lost at the connection portion. Moreover, area extraction processing 2
In the line image obtained by tracing the boundary line of the region extracted in, the smooth contour line (hereinafter,
Other than the original contour line), a projecting contour line is included, so that the projecting contour line needs to be removed.

FIG. 7 is an explanatory diagram of the recognition process for the circular line image performed in the contour line recognition process 4 of FIG.
Conventionally, in the contour line recognition process 4, it is assumed that the projection portion included in the contour line 10 is not connected to the target object region and the projecting region with a width of a certain length w or more, and the contour line 1
A point B on the contour line 10 at which the straight line distance from a point A on 0 is w and the length L of the contour line 10 between the point A and the point A is maximum is obtained. If the length L is larger than a preset threshold value, the contour line 10 between the points AB is detected / removed as a protruding portion, and the area between them is linearly interpolated.

[0009]

However, the conventional image recognition system has the following problems. Figure 8
(A), (b) is explanatory drawing of the conventional problem. As shown in FIG. 8A, in the contour line recognition processing 4 of FIG. 6, since the projection portion is interpolated by a straight line having a constant length w, the projection portion has a width shorter than w (point a-point b). When connecting with each other, a part (between points a and A) of the contour line of the protruding portion remains.

Further, as shown in FIG. 8B, if there is a portion of the contour line of the protruding portion that is close to the original contour line, interpolation is performed at the approaching portion (between points A and B). And then
A part of the original contour line (between point b and point B) is also removed.

As described above, in the conventional method, in addition to the original contour line to be obtained, a part of the contour line of the protruding portion may remain, or an interpolation straight line that does not reflect the original contour line may appear. However, in the shape recognition process performed based on the obtained contour line, it causes erroneous recognition, and it is difficult to solve them.

The present invention has the problem that the conventional technique has a problem that, in addition to the original contour line to be obtained, a part of the contour line of the protrusion remains, or the interpolation does not reflect the original contour line. An image recognition method which solves the problem that a straight line may appear is provided.

[0013]

In order to solve the above-mentioned problems, the present invention provides an image recognition method for recognizing an image of a circular line figure, in which data of a circular line figure including a protruding curved portion is stored. Input and use a polar coordinate system centered on the intersection of the diagonal lines of the circumscribed rectangle of the line figure given as the coordinate system,
A curved portion corresponding to the circular portion of the circular shape in the line figure is extracted. Then, from the extraction result of the curved portion,
By detecting and removing the protruding curved portions and interpolating between them, a circular line figure that does not include the protruding curved portions is obtained.

[0014]

According to the present invention, since the image recognition system is configured as described above, the process of extracting the curved line portion corresponding to the circular portion of the circular shape in the line figure is applied to the given protruding curve. It is possible to extract a curved line portion corresponding to a contour line resulting from the original circular shape from the circular line shape including the portion. From this extraction result, the portion of the original contour line becomes clear, and only the protruding portion is interpolated. This allows
Unnecessary projecting contour line portions included in a circular line figure containing noise are removed, and a smooth circular line figure reflecting the original contour line can be extracted. Therefore, the above problem can be solved.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a diagram for explaining the processing contents of an image recognition method for a circular line image showing an embodiment of the present invention. In this image recognition method, grayscale image data is input as an original image by image input processing 1 as shown in FIG. 6 of the related art, and is sent to area extraction processing 2. When the shape of the target object is extracted as an area on the image from the original image by the area extraction processing 2, a part of the contour line extraction processing 3 is executed on the extraction result, and the execution result is the coordinates of FIG. It is input to the center determination processing 11.

In the coordinate center determination processing 11, the center of the polar coordinate system is determined with respect to the input contour line data, and the determination result is sent to the polar coordinate system conversion processing 12. In the polar coordinate system conversion process 12, the given circular line figure is converted into the polar coordinate system, and the conversion result is sent to the maximum distance point extraction process 13. In the distance maximum point extraction processing 13, the point having the maximum distance viewed from the center in each direction is extracted from the line figure converted into the polar coordinate system, and further smooth section extraction processing 14 is performed to extract a section that does not include a protruding portion. Then, the obtained section is sent to the protrusion portion determination processing 15. In the projection portion determination processing 15, the projection portion is determined from the input section, and the projection portion is linearly interpolated by the projection portion interpolation processing 16 to remove the projection portion.

These processes 11 to 16 are executed by a program process using a computer. Next, FIG.
~ Each processing content in Fig. 1 will be described in detail with reference to Fig. 5.

FIG. 2 is a polar coordinate system conversion process 1 in FIG.
2, FIG. 3 is a diagram showing the maximum distance point extraction process 13, FIG. 4 is a diagram showing a smooth section extraction process 14, and FIG.

In the image recognition method of this embodiment, as shown in FIG. 2, paying attention to the fact that the target line figure, that is, the line image 20 has a circular shape, and the coordinate center determination processing 11 causes the line figure 20 to be displayed. The intersection of the diagonal lines in the circumscribed quadrangle 21 is the center O of the coordinate system. Then, the polar coordinate system conversion process 1
2, the coordinate system is converted into the polar coordinate system (r, θ) using the angle θ, the radius r, and the coordinate s, and the conversion result is sent to the maximum distance point extraction processing 13 shown in FIG.

As shown in FIG. 3, maximum distance point extraction processing 1
3, when the line image 20 has a protrusion, a plurality of points on the line image 20 having coordinate values of a certain angle θ may appear as points r1, r2, and r3. Here, if it is assumed that the protrusion portion only occurs inside (or outside) of the circumferential portion of the circular line image 20 when viewed from the center O, then the point from the center O among these points r1, r2, and r3. The point with the maximum (or minimum) distance r is the most promising candidate as the point on the original contour line.

The following is a description of the processing contents in the case where it is assumed that the protruding portion is generated only inside the circumferential portion. It should be noted that in the following processing, the same processing contents are obtained even when the protruding portion is generated outside. As shown in FIG.
In the maximum distance point extraction processing 13, the point where the distance r from the center O is the maximum when viewed in the direction of each angle θ from the center O, and along the curve of the line image 20 given that point. The coordinate s representing the length is obtained, and the result of maximum distance point extraction is sent to the smooth section extraction processing 14 shown in FIG.

In FIG. 4, a thick line in the line image 20 is a smooth section 20a, a thin line is a protruding portion 20b, and Δθ is an angle θ.
, And Δs is the minute change amount of the coordinate s. Figure 4
As shown in FIG.
The ratio Δs / Δθ is obtained from the minute change amount Δs of the coordinate s with respect to the minute change amount Δθ of. If the given line image 20 is a perfect circle having a radius γ0, the value of this ratio Δs / Δθ is γ0. In the line image 20, a protruding curved portion (that is,
When the protrusion portion) 20b is included, the value of the ratio Δs / Δθ becomes much larger than γ0. Here, the section where the ratio Δs / Δθ continues on the given circular line image 20 with a value of about γ0 which is assumed in advance is called a smooth section 20a.

The extraction result of the smooth section extraction processing 14 is
It is sent to the projection portion determination processing 15 and the projection portion interpolation processing 16.

In FIG. 5, a broken line curve 20b is a protrusion, and a thin line 21 between t1 and t2 is an interpolation straight line. As shown in this figure, the smooth section 20a obtained by the smooth section extraction processing 14 is a section in which the original circular contour line is correctly reflected in the given line image 20. The curved portion in the line image 20 which is not included in the smooth section 20a is caused by the protruding portion 20b.

Therefore, in the projection portion determination processing 15, the point t
At both end points of a straight line portion not included in the smooth section 20a, such as 1 and t2, it is determined whether or not the line segment connecting these two points sufficiently reflects the original contour shape of the circular figure. The determination is made based on the polar coordinate values of the two points t1 and t2. Then, using this determination result, the projection portion interpolation processing 16 interpolates between the points t1 and t2 with the interpolation straight line 21.

By the above processing, the smooth section 20a and the interpolation straight line 21 are obtained from the circular line image 20 including the given projecting curve portion (projection portion) 20b, which is an unnecessary noise component. It is possible to obtain a circular line drawing that correctly reflects the original circular shape that does not include the protruding portion 21b.

In the above embodiment, the method for recognizing a circular line figure such as the crimp head of a rivet has been described, but the present invention is also applicable to a method for recognizing a circular line figure other than the rivet. it can.

[0028]

As described above in detail, according to the present invention, of the circular line figures including the given projecting curved portions, the contour lines are derived from the original circular shape. All the curved line parts can be extracted by the operation of obtaining a smooth section. As a result, the part of the original contour line lost due to the connection of the protruding contour lines becomes clear,
By interpolating only this portion, it is possible to avoid the problem that the contour line of the protruding portion remains in the conventional method and that the original contour line is removed. Therefore, it is possible to accurately obtain a circular line figure that accurately reflects the original circular figure. Therefore, it can be applied to various inspection systems such as rivet defect detection.

[Brief description of drawings]

FIG. 1 is a diagram illustrating processing contents of an image recognition method according to an embodiment of the present invention.

FIG. 2 is a diagram showing a polar coordinate system conversion process 12 in FIG.

FIG. 3 is a diagram showing a maximum distance point extraction process 13 in FIG. 1.

FIG. 4 is a diagram showing a smooth section extraction process 14 in FIG.

5 is a diagram showing a projection portion interpolation processing 16 in FIG.

FIG. 6 is a diagram showing processing contents of a conventional image recognition method.

FIG. 7 is an explanatory diagram of circular line image recognition processing in FIG.

FIG. 8 is an explanatory diagram of a conventional problem.

[Explanation of symbols]

 11 Coordinate Center Determination Processing 12 Polar Coordinate System Conversion Processing 13 Distance Maximum Point Extraction Processing 14 Smooth Section Extraction Processing 15 Protrusion Part Determination Processing 16 Protrusion Part Interpolation Processing

Front Page Continuation (72) Inventor Hiroyuki Terakasa 10 Oemachi, Minato-ku, Nagoya, Aichi Prefecture Mitsubishi Heavy Industries, Ltd. Nagoya Aerospace Systems Works

Claims (1)

[Claims] 1. A polar coordinate system centered on an intersection of diagonal lines of a circumscribed quadrangle of a line figure given as a coordinate system, by inputting data of a circular line figure including a protruding curved portion,
By extracting a curved portion corresponding to the circular portion of the circular shape in the line figure, by detecting and removing the protruding curved portion from the extraction result of the curved portion and interpolating between them,
An image recognition method characterized in that a circular line figure that does not include the protruding curved portion is obtained.