JPH08271435A - Method for detecting push-in scratch of wheel rim for automobile - Google Patents

Abstract

PURPOSE: To detect a push-in scratch accurately while discriminating automatically from disturbance, e.g. a flaw, by irradiating the surface of a wheel rim of automobile obliquely from the opposite sides in the axial direction of rim, picking up the image of the surface of rim from the direction perpendicular to the axis of rim by means of a CCD camera, and processing the image where the edge of scratch exhibits high illuminance by means of a CPU. CONSTITUTION: A wheel rim 1 is irradiated, on the surface with light obliquely from the opposite sides in the axial direction of rim. Inner and outer circumferential surfaces of the rim are irradiated with light emitted from two light sources 6 respectively. Since the rim is irradiated obliquely, only the edge 3 of a push-in scratch 2 is rendered conspicuous in terms of luminance or density. Since the rim is irradiated with light from the opposite sides, the edge 3 on an arc is rendered conspicuous. Image of the rim surface is then picked up by means of a CCD camera from the direction perpendicular to the axis of rim and the image is processed by means of a CPU according to a detection algorithm. Since the CPU detects only the pull-in scratch while discriminating from disturbance, detection is automated and high speed processing is realized without requiring any skill.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting indentation damage on a wheel rim for an automobile.

[0002]

2. Description of the Related Art A disc wheel for an automobile comprises a rim and a disc welded and joined together. The rim is manufactured by rolling a steel sheet and butt-welding the ends to form a cylindrical body, which is then formed into a rim shape by a multistage roll. When butt welding, the end of the rolled steel plate is gripped by a chuck, and flash butt welding is applied along the butt line.In this case, foreign matter during welding easily adheres to the chuck, and the chucked steel sheet adheres the foreign matter. If you grip the end of the, the rim will be pushed (dented) and scratched.
Unlike ordinary scratches, the indentation scratches are likely to cause coating unevenness and coating defects in the coating of the final wheel manufacturing process. In addition, indentation scratches are not desirable in appearance. Further, if the indentation is on the bead seat portion (contact surface with the tire), air leakage may occur. For this reason, conventionally, a visual inspection is performed by visual judgment after manufacturing the rim and before welding and joining with the disc.

[0003]

However, the conventional visual judgment has the following problems. Since the line speed is high and 100% inspection is required, the workload of the inspectors is high and requires multiple personnel. In addition, since the judgment is made by visual inspection, there is a possibility that it may be overlooked due to differences in the wound site, size, shape, etc., and high skill is required for the inspector. In the method that uses image processing software that uses a general-purpose CCD camera, indentation scratches are generated by chucks in the welding process, so there are heat-affected zones (color unevenness) and scratches in the welded portion at the inspection location. It is difficult to detect an indentation flaw by distinguishing it from those noises. Further, when inspecting the curvature of the rim and both inside and outside, it is difficult to reliably inspect the indentation damage. Other methods using various sensors have problems such as measurement pitch, dead zone, incidental equipment, measurement time, cost, etc., and currently only relying on an inspector. An object of the present invention is to provide an automobile wheel capable of automatically detecting scratches and accurately detecting indentation scratches by distinguishing them from disturbances (noise) such as heat-affected zones (color unevenness) and scratches of welded portions. It is to provide a method for detecting a rim indentation flaw.

[0004]

The method of the present invention for achieving the above object of the present invention is as follows. The step of illuminating the surface of the wheel rim for automobiles obliquely from both sides in the axial direction of the wheel rim, and capturing the image of the wheel rim surface with a high brightness edge of a scratch by a CCD camera from a direction perpendicular to the axial direction of the wheel rim, and the captured wheel Image processing of the rim surface image by a CPU, wherein the image processing step divides the image into n pixels in the rim axis direction and m pixels in the rim radial direction to form adjacent pixels in the rim axis direction. Calculating the relative density of all the pixels, and the image has i (where i <n) pixels in the rim axis direction and j (where j <m) pixels in the rim radial direction. The detection area is divided into one detection area, and the detection area edge density is calculated for the detection area to obtain A% (where A is 6 to 1).
It is determined whether or not it is equal to or more than a value set to 0), and the edge aspect ratio in the detection area is calculated for the detection area, and the edge aspect ratio in the detection area is B (where B is 0.
(Value set to 4 to 0.8) or more, and a detection area having an edge density in the detection area of A% or more and an edge aspect ratio in the detection area of B or more is selected, and the selection detection is performed. For the area only, the detection area is i in the rim axis direction.
It is divided into a plurality of (where i ′ <i) and j subdivided detection areas having j pixels in the rim radial direction, and one of the subdivided detection areas has an edge aspect ratio within the subdivided detection area. The edge density in the subdivision detection area is calculated to leave only edges having a ratio of C or more (where C is set to 0.2 to 0.4), and the edge density in the subdivision detection area is calculated. The edge density in the subdivision detection area is D%
A step of removing edges less than (where D is set to 1 to 5) for all subdivision detection areas of the selection detection area, and there is an edge matching the conditions of the step It is assumed that the edge is an edge of an indentation, and otherwise, it is a disturbance, and the process of moving to the next detection area and executing the steps after the detection area selection step are repeated for all the detection areas. An indentation detection method for an automobile wheel rim, which comprises:

[0005]

In the above-described method of the present invention, the edge of the indentation flaw is raised by applying light, the image is captured by the CCD camera, and the indentation injury is discriminated from the disturbance by the CPU of the computer. No matter what, it happens automatically. Also, since the light is directed from both sides in the axial direction of the wheel rim, both the upper edge and the lower edge of the indentation scratches that are almost circular and have a certain size float up at the same time, and scratches that extend almost in a straight line. Make it easy to distinguish. Further, in the image processing step, since the relative density with the adjacent pixel is calculated in the rim axis direction, the influence of the color unevenness of the weld portion extending in the rim axis direction is automatically removed, and the direction orthogonal to or intersecting with the rim axis direction. Indented scratch edges and scratched pixels can be selected. In image processing,
The image is divided into a plurality of detection areas, and the detection area whose edge density in the detection area is A% or more and whose edge aspect ratio in the detection area is B or more is picked up. You can selectively pick up multiple parallel scratches similar to indented scratches. Further, in the image processing, only the detection area picked up above is further subdivided so that the operation for distinguishing the indentation scratch and the scratch is performed, so compared to the case of performing the operation for all the detection areas, Judgment time can be greatly reduced and real-time online detection is possible. In the image processing, the picked-up detection area is subdivided, and the subdivided detection area is
Edges in the subdivision detection area The aspect ratio of C or more is picked up and the edges with the edge density in the subdivision detection area of less than D% are removed, so that the edge of the indentation and the parallel scratch can be distinguished. , Only the edge of indentation can be detected. Thus, it is possible to detect the indentation with high accuracy, which is distinguished from the disturbance.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIGS. 1 to 3 show an image capturing process of a method according to an embodiment of the present invention. As shown in FIG. 1, light is radiated obliquely from both sides in the axial direction of the wheel rim onto the surface of a vehicle wheel rim 1 (cylindrical one being manufactured). Reference numeral 6 denotes a light source that emits light in the form of a beam, and two light sources that illuminate the outer peripheral surface of the wheel rim and two light sources that illuminate the inner peripheral surface of the wheel rim are provided. The diagonal angle is 10 to 10 from the rim surface
It is set to 40 °. By shining light diagonally,
As shown in FIG. 3, only the edge 3 of the indentation scratch 2 stands out in comparison with other portions in terms of brightness or density. Since light is applied from both sides in the rim axis direction, a total of two or more arc-shaped edges 3 stand out at the upper and lower opposing portions of the circle in FIG. The surface of the wheel rim is picked up by the CCD camera 4 from the direction perpendicular to the axial direction of the wheel rim, and an image is captured. At least one CCD camera 4 (two in the figure) is arranged on the outer side and the inner side of the wheel rim 1. Since the inner and outer surfaces are imaged, it is possible to detect scratches on the inner surface side of the wheel rim, which were difficult with conventional visual inspection. Then, this image, C
Image processing is performed by the CPU of the computer connected to the CD camera 4.

Image processing in the CPU is executed according to the detection algorithm shown in FIG. In step 101, raw image data is input. Then, in step 102,
The image is divided into n pixels in the rim axis direction (corresponding to the vertical direction of the image) and m pixels in the rim radial direction (corresponding to the horizontal direction of the image). In this case, it is desirable that n is 480 and m is 512, or that it has more pixels. Then, the calculation of the relative density with respect to the adjacent pixels in the rim axis direction is executed for all the pixels. As shown in FIG. 5, the pixels are (1, 1), (2, 1), ...
When the numbers (1, 1), (1, 2), ... Are given in the horizontal direction of the image, the relative density of the pixel of (1, 1) is, for example,
The density of the pixel of (1, 1) -the absolute value of the density of the pixel of (2, 1). Since the relative densities of adjacent pixels are calculated in the vertical direction of the image, the color unevenness of the welded portion 5 extending in the wheel rim axial direction is automatically canceled, and the disturbance (noise) due to the color unevenness can be removed.

Then, in step 103, step 10
The image divided into pixels in 2 is divided into a plurality of detection areas having i (where i <n) pixels in the rim axis direction and j (where j <m) pixels in the rim radial direction. For example, i is 16 and j is 16. In this case, one detection area has i × j pixels. Then, one detection area for calculation is set. Then, for the set detection area, the calculation from the next step 104 to step 112 is executed. In step 104, the edge density in the detection area (percentage of the total number of pixels in the detection area out of the total number of pixels in the detection area) is calculated, and the edge density in the detection area is a predetermined value A% (however, A is a value set to 6 to 10, and it is determined whether it is 8) or more, for example. In step 105, the edge aspect ratio in the detection area (the ratio of the vertical length Y to the horizontal length X of a group of a plurality of pixels having an edge in the detection area,
6 and 7) to calculate whether the edge density in the detection area is a predetermined value B (where B is a value set to 0.4 to 0.8, for example, 0.6) or more. To determine. An edge density in the detection area of A% or more and an edge aspect ratio in the detection area of B or more are selected as a detection area having an indentation scratch or a parallel scratch similar thereto. For example, a scratch as shown in FIG. 6 is splashed because the aspect ratio Y / X is 0.4 or less, and an indentation as shown in FIG. 7 and a parallel scratch as shown in FIG. 8 have an aspect ratio Y / X. Since it is 0.6 or more, it is judged to be an indentation scratch or a similar scratch.

Next, in steps 104 and 105, the density in the detection area is A% and the edge density in the detection area is A%.
As described above, if the detection area with the edge aspect ratio in the detection area of B or more is not selected, the process jumps to step 110. If the detection area is selected, the process proceeds to steps 106 to 108 for only the selected detection area, and parallel scratches with indentation scratches are made. An operation for distinguishing from a disturbance such as a scratch is executed. Step 106
Then, as shown in FIG. 9, the detection area is divided into a plurality (for example, 4) in the rim axis direction (longitudinal direction) (the rim radial direction is unchanged), and i ′ pieces (however, i ′ are arranged in the rim axis direction). <
i and i ′ are, for example, 4), and are divided into a plurality of subdivided detection areas having j pixels in the rim radial direction. Next, in step 107, the edge aspect ratio in the subdivision detection area is calculated for one of the subdivision detection areas, and the edge aspect ratio in the subdivision detection area is C (where C is 0.2).
Leave only edges of ~ 0.4, eg 0.3) (C and below are discarded as scratches). Then, step 10
In step 8, the edge density in the subdivision detection area is calculated, and the edge density in the subdivision detection area is D% (where D is 1 to
Edges less than 5, eg 2) are discarded as scratches. The above steps 106 to 108 are executed for all subdivided detection areas (four areas in the above example) of the selection detection area, and thereby, among the scratches in the selection detection area, the scratches similar to the indentation scratches are removed, and the indentation scratches are removed. Only elect. Then, in step 109, the edge aspect ratio in the detection area is recalculated to confirm that the aspect ratio is B (eg, 0.6) or more. Step 109 may be skipped.

Then, in step 110, step 1
It is determined whether or not there is an edge that matches the process conditions of 03 to 109 (if there is, flag ic = 0), and if there is, the edge is taken as the edge of the indentation scratch, and if not, step 112. To be a disturbance (noise). Next, in step 113, the detection area set in step 103 is moved to the next detection area, and the next detection area is newly set as the set detection area in step 10
Performing 3 to 113 is repeated for all the detection areas. When the determination of the indentation damage on all the detection areas is completed, the determination of the indentation damage of the wheel rim is completed, and the process proceeds to the determination of the next wheel rim.

[0011]

According to the present invention, the edge of the scratch is highlighted by shining light, and the image is captured by the CCD camera, and the C
Since only the indentation is detected by the PU by distinguishing it from the disturbance,
The detection is automated without visual inspection, requires no skill, and can be processed at high speed (for example, within 1 second). Also,
Since the light is applied from both sides in the axial direction of the wheel rim, both the upper edge and the lower edge of the indentation can be read at the same time, the aspect ratio of the edge is large, and it is easy to distinguish from the scratch with a small aspect ratio. There is. Further, in the image processing step, since the relative density with respect to the adjacent pixel is calculated in the rim axis direction, the disturbance of the color unevenness of the welded portion can be automatically removed. Further, since the detection area having the edge density in the detection area of A% or more and the edge aspect ratio in the detection area of B or more is selected, only the indentation flaw and the flaw similar thereto can be picked up. Moreover, since the picked-up detection area is further subdivided and examined, the indentation can be distinguished from the scratch. Thus, the indentation can be detected with high accuracy without being disturbed.

[Brief description of drawings]

1 is a perspective view of an apparatus for carrying out the method of an embodiment of the present invention.

FIG. 2 is a sectional view of a detection edge.

FIG. 3 is a front view of a detection edge.

FIG. 4 is a flowchart showing a detection algorithm.

FIG. 5 is a front view of an image.

FIG. 6 is a relationship diagram between a side surface of a work and a front surface of a detection area in the case of a scratch.

FIG. 7 is a relationship diagram between a side surface of a work and a front surface of a detection area in the case of an indentation scratch.

FIG. 8 is a relationship diagram between a side surface of a work and a front surface of a detection area in the case of parallel scratches.

FIG. 9 is a front view of a state in which a detection area is divided into a plurality of subdivided detection areas in the case of parallel scratches.

[Explanation of symbols]

 1 Wheel rim 2 Indentation scratch 3 Edge 4 CCD camera 5 Welding part 6 Light source

Front page continuation (72) Inventor Akihiko Yamada 9 Topy Kogyo Co., Ltd. 5-4 Yonbancho, Chiyoda-ku, Tokyo

Claims (1)

[Claims] 1. A step of irradiating the surface of an automobile wheel rim diagonally from both sides in the axial direction of the wheel rim, and capturing a wheel rim surface image with a high brightness of a scratch edge by a CCD camera from a direction perpendicular to the axial direction of the wheel rim. Image processing of the captured wheel rim surface image with a CPU, wherein the image processing step divides the image into n pixels in the rim axis direction and m pixels in the rim radial direction to separate the rim axis. Calculating relative densities with respect to adjacent pixels in all directions, i images in the rim axis direction (where i <n) and j images in the rim radial direction (where j <m) ) Is divided into detection areas each having one pixel, and one detection area for performing calculation is set, and the edge density in the detection area is calculated for the detection area, and the edge density in the detection area is A% (only , A is determined whether or not the setting value) or more of the 6-10, and the detection area edge aspect ratios and calculates the detection area edge aspect ratio for the detection area B (however, B
Is a value set to 0.4 to 0.8) or more, and selects a detection area in which the edge density in the detection area is A% or more and the edge aspect ratio in the detection area is B or more, With respect to only the selected detection area, the detection area is divided into a plurality of subdivided detection areas having i ′ pixels (where i ′ <i) in the rim axis direction and j pixels in the rim radial direction, and the subdivided detection areas are detected. The edge aspect ratio within the subdivision detection area is calculated for one of the areas, and the edge aspect ratio within the subdivision detection area is C (where C is set to 0.2 to 0.4) or more. Leaving edges, and calculating the edge density in the subdivision detection area to remove edges whose edge density in the subdivision detection area is less than D% (where D is set to 1 to 5). The total subdivision detection area of the selection detection area Steps to be executed by the following, and if there is an edge that meets the conditions of the above step, the edge is considered to be the edge of the indentation, otherwise the step is determined to be disturbance, and the process moves to the next detection area and the detection area is selected. A method for detecting indentation damage of a wheel rim for an automobile, comprising: performing the steps after the step for all the detection areas.