Inspection equipment and inspection method

Abstract

To provide an inspection device that can perform inspection as precisely as possible even if the brightness of lighting emitted to a material to be inspected is not uniform.SOLUTION: A CPU 140 combines a predetermined number of reference images not including a welded part (S13), averages photographed images after the combination (S14), averages brightness values of pixels in the reference images after the averaging in a y-axis direction (S15), approximates the brightness value (brightness projection histogram) after the averaging by the n-th order polynomial (S16), calculates an inverse number of the approximated brightness value and takes it as a brightness correction coefficient (S17). The CPU 140 corrects an evaluation image by using the brightness correction coefficient and performs image recognition on the evaluation image after the correction, and thereby detects the welded part.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] This application relates to an inspection apparatus and an inspection method for inspecting a welded part of a material.

Background Art

[0002] Patent Document 1 describes a printed paper inspection apparatus that detects the amount of reflected light for one sheet per pixel, calculates the maximum value of the amount of reflected light detected for each line for each line, and corrects the detected amount of reflected light using this maximum value. In this inspection apparatus, the corrected amount of reflected light is compared with a reference value to detect whether there are printing defects such as stains on the printed paper.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the inspection apparatus described in Patent Document 1, it is assumed that the luminance of the illumination irradiated on the test material is uniform over the entire test material, and non-uniformity is not considered. Therefore, when the luminance of the illumination irradiated on the test material is non-uniform, there is a risk that the occurrence of printing defects cannot be correctly detected.

[0005] An object of this application is to provide an inspection apparatus and an inspection method that can perform the inspection as correctly as possible even when the luminance of the illumination irradiated on the test material is non-uniform.

Means for Solving the Problems

[0006] To achieve the above objective, the inspection apparatus of the present invention comprises a camera for photographing the material to be inspected, an illumination device for illuminating the camera's shooting range with a second optical axis in a direction different from the camera's first optical axis, and a control processing device that uses the captured image obtained by the camera as an evaluation image, performs image processing on the evaluation image, and determines whether or not the material to be inspected contains any abnormal parts, wherein the control processing device uses one or more captured images obtained by photographing the material to be inspected that does not contain any abnormal parts as reference images, and generates a luminance projection histogram by averaging the luminance value of each pixel of the reference image over an axis orthogonal to the axis generated by projecting the second optical axis onto the image plane of the reference image, and then processes the luminance projection histogram generated by the first averaging process in order of n (where n is 3 The following steps are performed: first, the luminance projection histogram is approximated by an approximation function composed of the above integer polynomials, then the reciprocal of the approximated luminance projection histogram is calculated, or the reciprocal of the luminance projection histogram is calculated and then approximated by the approximation function to obtain the luminance correction value; and second, the evaluation image acquired by camera photography is corrected by the luminance correction value obtained by the approximation function. In the image processing, image processing is performed on the evaluation image corrected by the correction process to determine whether or not the material under inspection contains any abnormal parts.

[0007] To achieve the above objective, the present invention provides an inspection method that uses a camera to photograph the material to be inspected and an illumination device to illuminate the camera's shooting range with a second optical axis in a direction different from the camera's first optical axis, uses the captured image obtained by the camera as an evaluation image, performs image processing on the evaluation image to determine whether or not the material to be inspected contains abnormal parts, and uses one or more captured images obtained by photographing the material to be inspected that does not contain abnormal parts as reference images, and generates a luminance projection histogram by averaging the luminance values ​​of each pixel of the reference image over an axis orthogonal to the axis generated by projecting the second optical axis onto the image plane of the reference image, and then processes the luminance projection histogram generated by the first averaging step to the nth order (where n is 3The process includes a calculation step of approximating the luminance projection histogram with an approximation function composed of the above integer polynomials, then calculating the reciprocal of the approximated luminance projection histogram, or calculating the reciprocal of the luminance projection histogram and then approximating it with the approximation function to obtain a luminance correction value, and a correction step of correcting the evaluation image acquired by camera shooting with the luminance correction value obtained in the calculation step, and in the image processing, image processing is performed on the evaluation image after correction in the correction step to determine whether or not the material under inspection contains abnormal parts. [Effects of the Invention]

[0008] According to this invention, even if the brightness of the image of the material to be inspected, illuminated by the light source, is uneven, it becomes possible to perform the inspection as accurately as possible. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows a schematic configuration of a molding system including an inspection device according to one embodiment of the present invention. [Figure 2] This figure shows an example of the arrangement of the camera and lighting fixtures included in the inspection device shown in Figure 1. [Figure 3] This block diagram shows the schematic configuration of the PC that controls the molding system shown in Figure 1. [Figure 4] Figure 1 is a flowchart showing the steps involved in the brightness correction coefficient generation process performed by the PC, particularly the CPU. [Figure 5] Figure 1 is a flowchart showing the procedure for the weld detection process performed by the PC, specifically the CPU. [Figure 6] Figure 5 shows an example of evaluation images before and after correction by the correction process within the weld detection process. [Figure 7] This figure shows an example of a photograph that includes a welded area. [Figure 8] Figure 1 is a flowchart showing the steps involved in the image cropping process performed by the PC, specifically the CPU. [Figure 9] Figure 8 is a diagram illustrating the trimming threshold determination process within the image trimming process. [Figure 10] This figure compares the average error between the luminance projection histogram in the y-axis direction and the luminance approximated by a first- to ninth-order polynomial for each company's material coils. [Modes for carrying out the invention]

[0010] The embodiments of this application will be described in detail below with reference to the drawings.

[0011] As shown in Figure 1, a molding system 1 including an inspection device 20 (an example of an "inspection device") according to one embodiment of the present invention is mainly composed of a powered material table 10, an inspection device 20, a material straightening roller 30, a material feeder 40, and a molding machine 50, extending from the upstream side to the downstream side where the material coil 200 (an example of a "material to be inspected"), which is the material to be molded, is transported.

[0012] In this embodiment, the material coil 200 is a roll-shaped coil produced by winding a plurality of plate-shaped coils of a predetermined length, connected longitudinally to extend the overall length. In this embodiment, the longitudinal connections of each plate-shaped coil are made by welding. This roll-shaped material coil 200 is placed on a powered material stand 10, and when the molding system 1 is instructed to start molding, it is unfolded by the powered material stand 10 and sent downstream.

[0013] The inspection device 20 is a device that detects welding locations 201 (see Figure 2; an example of an "abnormal location") by inspecting the surface of the material coil 200 being conveyed in the downstream direction. As shown in Figure 2, the inspection device 20 is equipped with a camera 21 and a lighting fixture 22. The lighting fixture 22 illuminates the material coil 200 with light, and the camera 21 photographs the illuminated material coil 200. In Figure 2, the vertical upward direction is indicated as "up," the vertical downward direction as "down," the horizontal right direction as "front," and the horizontal left direction as "back." Also in Figure 2, the front of the page is considered the right, and the back of the page as the left.

[0014] The camera 21 is installed above the material coil 200 and photographs the material coil 200. The photographed image obtained by the camera 21 is transmitted to a PC 100 (included in the "inspection device") described later using FIG. 3.

[0015] In this embodiment, the lighting fixture 22 employs LED bar lighting in which LED elements are arranged in a bar shape and is installed below the front side of the camera 21. That is, the lighting fixture 22 is installed in a direction in which its optical axis 22a (an example of the "second optical axis") is different from the optical axis 21a (an example of the "first optical axis") of the camera 21. Therefore, the lighting fixture 22 illuminates the imaging range photographed by the camera 21 from obliquely above. Ideally, the imaging range should be illuminated from directly above by the lighting fixture. When the imaging range is illuminated from obliquely above as in this embodiment, unevenness in the luminance of the irradiation light occurs on the imaging range, and as a result, unevenness in the luminance also occurs in the photographed image. However, due to reasons such as wanting to make the inspection device 20 compact, wanting to manufacture the inspection device 20 at a low cost, and making the lighting fixture 22 illuminate as efficiently as possible even with the minimum necessary lighting, and crossing the respective optical axes of the camera 21 and the lighting fixture 22 on the surface of the material coil, the lighting fixture 22 has to be installed at a position where it illuminates the imaging range from obliquely. Note that the position where the lighting fixture 22 is installed is not limited to the lower front side of the camera 21 and may be the lower rear side.

[0016] Returning to FIG. 1, the material correction roller 30 corrects the material coil 200 being conveyed so that it becomes flat, and the material feeder 40 sends out the corrected material coil 200 to the molding machine 50.

[0017] In this embodiment, the molding machine 50 mainly performs a pressing process of manufacturing a plate material of a predetermined shape by pressing the material coil 200 sent out from the material feeder 40 and bending the plate material into a predetermined shape to manufacture a product, and a sorting process of sorting the products manufactured in the pressing process into good products and defective products.

[0018] As described above, the material coil 200 is constructed by connecting multiple plate-shaped coils by welding to extend its overall length. This is because there is a demand to reduce the number of times the material coil 200 is replaced during the product manufacturing process, thereby reducing the downtime associated with replacing the material coil 200. Since the length of a plate-shaped coil is determined during the manufacturing process of that coil, this demand is met by extending it by welding multiple plate-shaped coils together. Consequently, the material coil 200 contains multiple welded areas. However, if these welded areas are included in the final product, the product's functionality may not be satisfied. Therefore, it is necessary to ensure that the welded areas are not included in the final product.

[0019] To address this, conventional molding systems use material coils with markings near each welded area, and during the manufacturing process, optical sensors are used to detect the markings on the material coils. When a marking is detected, a rejection signal is output to the molding machine 50. Upon receiving the rejection signal, the molding machine 50 separates a predetermined number of products before and after the product that is expected to contain a welded area as defective products in the sorting process described above.

[0020] However, marking the vicinity of the welded area when winding a welded plate coil increases the cost of the material coil. Therefore, there has been a need for the development of a method to detect the welded area even when using an unmarked plate coil.

[0021] The molding system 1 of this embodiment uses a material coil 200 that is not marked near the weld area, and detects the weld area contained in the material coil 200 during the product manufacturing process. When a weld area is detected, it outputs a rejection signal to the molding machine 50.

[0022] Figure 3 shows the control configuration of PC100, which controls the molding system 1. As shown in Figure 3, PC100 includes a user IF110, a display 120, an input / output IF130, a CPU 140, a memory 150, and a communication IF160. The user IF110, display 120, CPU 140, memory 150, and communication IF160 are interconnected via a bus 170. IF is an abbreviation for interface.

[0023] The user interface 110 typically consists of a keyboard and mouse. The display 120 consists of a display device such as a liquid crystal display or an organic EL display, and a drive circuit that drives the display device. If a touch panel type is used as the display 120, the user can perform input operations by touching input buttons on the screen. Therefore, in this case, the display 120 also serves as the user interface 110.

[0024] In this embodiment, the input / output IF130 uses USB, but it is not limited to this and may also use Bluetooth® or NFC. The camera 21 is connected to the input / output IF130.

[0025] The CPU 140 executes various application programs (hereinafter abbreviated as "apps"), including control processing programs, as well as firmware and the like.

[0026] Memory 150 includes ROM, RAM, HDD, SSD, and optical disc drive. The data storage area 150a of memory 150 is the area where the CPU 140 stores data necessary for executing control processing programs, etc. The program storage area 150b of memory 150 is the area where the OS, control processing programs, and various other applications and firmware are stored.

[0027] The communication IF160 connects the PC100 to the communication network 300. The communication network 300 is typically a wired or wireless LAN, but is not limited to these. The communication IF160 can communicate with the molding machine 50 via the communication network 300.

[0028] Figure 4 shows the procedure for generating the brightness correction coefficient, which is performed by the PC100, and specifically by the CPU140. The brightness correction coefficient generation procedure is performed before the weld detection procedure, which will be described later using Figure 5. Hereafter, in the explanation of each procedure, the step will be denoted as "S".

[0029] In Figure 4, the CPU 140 first photographs the non-welded portion of the material to be molded, i.e., the material coil 200 (S11). More specifically, the material coil 200 is attached to the powered material stand 10, and the leading edge of the material coil 200 is fed downstream by the material feeder 40 via the material straightening roller 30 so that the non-welded portion of the material coil 200 is within the shooting range of the camera 21. Then, the CPU 140 issues a shooting instruction to the camera 21. In response, the camera 21 photographs the shooting range and outputs the acquired image as a reference image to the input / output IF 130 of the PC 100. The reference image input from the input / output IF 130 is subjected to the image cropping process described later using Figure 8 and temporarily stored in the data storage area 150a of the memory 150.

[0030] Figure 6(a) shows an example of a reference image obtained by photographing a non-welded area on the material coil 200. Figure 6(b) shows a luminance projection histogram and its approximation. Figure 6 will also be used to show the before and after correction of the evaluation image described later using Figure 5, but for convenience, it will also be used to explain the luminance correction coefficient generation process. The reference image shown in Figure 6 is grayscale bitmap data, that is, data in which the luminance of one pixel is represented by an 8-bit numerical value (any integer value between 0 and 255). In Figure 6, the vertical upward direction is the positive direction of the y-axis (an example of "an axis orthogonal to the axis generated by projecting the second optical axis onto the image plane of the reference image" and "width direction"), and the horizontal right direction is the positive direction of the x-axis. The positive direction of the y-axis corresponds to the left direction in Figure 2, and the positive direction of the x-axis corresponds to the forward direction in Figure 2.

[0031] Returning to Figure 4, the CPU 140 then determines whether a predetermined number of images have been captured (S12). Here, the predetermined number is, for example, three images. If the determination in S12 indicates that the predetermined number of images have not been captured (S12: NO), the CPU 140 returns to processing S11 and instructs the camera 21 to take images again, capturing the same area of ​​the material coil 200. The reference image acquired through the capture is sent from the camera 21 via the input / output IF 130 as described above, subjected to the image cropping process described later, and stored in the data storage area 150a of the memory 150.

[0032] On the other hand, in the judgment of S12, if the predetermined number of images have been captured (S12: YES), the CPU 140 synthesizes the predetermined number of reference images (S13) and averages the synthesized reference images (S14). In other words, in S13, the CPU 140 synthesizes the predetermined number of reference images by summing the brightness values ​​of each pixel over the predetermined number of images, and in S14, it averages the synthesized reference images by dividing the sum of each pixel value by the predetermined number of images (for example, 3 images). Here, if the predetermined number of images is A and the reference image captured on the mth time is Im, the average reference image I bar is calculated by the following equation (1).

[0033]

number

[0034] Thus, the synchronized addition of a predetermined number of reference images, obtained by photographing the same shooting range on the material coil 200 a predetermined number of times and then averaging them, is performed in order to reduce air noise contained in the reference images. Air noise, in this context, refers to the instability of the image due to fluctuations caused by air when shooting with air present between the object being photographed and the camera. The effect of reducing air noise increases as the number of images used for reduction increases, so the predetermined number may be more than the three images used in this embodiment. However, the more images used, the longer it takes to shoot and to process the images by compositing and averaging them, so the predetermined number should be determined by comparing the reduction effect with the time spent.

[0035] Next, the CPU 140 averages the brightness value of each pixel in the averaged reference image along the y-axis (S15). Here, assuming that the average reference image Ixy bar has B pixels along the y-axis, the average brightness value Lx bar in the y-axis direction for each pixel is calculated by the following equation (2). The reason for averaging along the y-axis in this way is that in the reference image shown in Figure 6(a), the brightness value of each pixel gradually changes along the x-axis (in this embodiment, the direction of the x-axis coincides with the direction of the axis projected onto the image plane of the reference image of the optical axis 22a of the lighting fixture 22). In other words, the brightness value of each pixel does not change along the y-axis. Needless to say, the axis of averaging is varied according to the direction in which the brightness value of each pixel changes. Also, even though the brightness value is averaged along the y-axis at the S15 stage, the effects of scratches and dirt that spread in the width direction, for example, remain, and will appear as noise in the brightness projection histogram described later. Refer to the brightness projection histogram shown in Figure 6(b).

[0036]

number

[0037] Next, the CPU 140 approximates the averaged luminance values ​​(hereinafter referred to as the "luminance projection histogram") with an n-th degree polynomial (where n is an integer of 2 or greater) (S16: see the approximation line shown in Figure 6(b)), calculates the reciprocal of the approximated luminance values ​​and uses it as the luminance correction coefficient (S17), and then terminates the luminance correction coefficient generation process. Alternatively, the averaged luminance values, i.e., the luminance projection histogram, may be inverted in the y-axis direction before approximation. The luminance correction coefficient Kx is calculated by the following equation (3). The luminance correction coefficient Kx is stored in the data storage area 150a of the memory 150. Furthermore, the reason for approximating the averaged luminance values, i.e., the luminance projection histogram, with an n-th degree polynomial in the S16 process is, similar to the S13 process, to reduce noise caused by scratches, dirt, etc., that are partially present on the material coil 200. In other words, in the weld detection process described later, we want to detect only the welded areas by removing only the brightness unevenness and representing the welded areas, scratches, and dirt as they are. Therefore, we want to extract as much of the brightness unevenness component as possible through approximation and find a brightness correction coefficient.

[0038] Furthermore, an n-th degree polynomial (where n is an integer greater than or equal to 2) can be generated by approximation calculations such as the least squares method. A degree of 2 or higher is preferable for the n-th degree polynomial because light from a luminaire spreads radially, and changes in brightness (non-uniformity of brightness) change exponentially. For practical purposes, a polynomial of degree 3 or higher is preferable. Figure 10 compares the average error between the brightness projection histogram in the y-axis direction and the brightness approximated by a 1st to 9th degree polynomial for each company's material coil. Approximation with a polynomial of degree 5 or higher is more preferable, as the average average brightness error for each company's materials falls within 1.0.

[0039]

number

[0040] Figure 5 shows the procedure for the weld detection process performed by the CPU 140. The weld detection process is performed after the completion of the brightness correction coefficient generation process and in response to the molding system 1 starting the molding process on the material coil 200.

[0041] In Figure 5, the CPU 140 first captures the current shooting range of the material coil 200, which is the material to be molded (S21). In S11 in Figure 4, the CPU 140 captures the area of ​​the material coil 200 that does not include the welded area while the material coil 200 is stopped, whereas in S21, the CPU 140 captures the area of ​​the material coil 200 that is within the shooting range at the time the shooting command is given, while the material coil 200 is being transported. Therefore, some of the evaluation images acquired by shooting in S21 may include the welded area, while others do not. The evaluation images are stored in the data storage area 150a of the memory 150, similar to the reference images acquired by shooting in S11.

[0042] Next, the CPU 140 reads the evaluation image acquired by the imaging in S21 from the data storage area 150a and corrects the read evaluation image using the brightness correction coefficient generated by the brightness correction coefficient generation process shown in Figure 4 above (S22). The corrected evaluation image I'xy is calculated by the following equation (4).

[0043]

number

[0044] Figure 6(a) shows a reference image used to explain the brightness correction coefficient generation process as described above. Figure 6(a) can also be described as the evaluation image acquired by the shooting in S21, that is, the evaluation image before the correction in S22. Figure 6(c) is the evaluation image after the correction in S22 has been applied to this evaluation image. In the evaluation image before correction, as shown in Figure 6(a), a brightness non-uniformity occurs where the brightness value gradually increases in the positive (forward) direction from the x-axis origin O. The cause of this brightness non-uniformity is that there is a misalignment between the optical axis 22a of the lighting fixture 22 and the optical axis 21a of the camera 21. In other words, the camera 21 photographs the material coil 200 from vertically above, while the lighting fixture 22 irradiates light to the shooting range of the camera 21 at a predetermined incident angle (>0°). Furthermore, as described above, it is not possible to position the lighting fixture 22 such that its optical axis 22a aligns with the optical axis 21a of the camera 21. Therefore, the above-mentioned brightness unevenness always occurs, albeit to varying degrees. In this embodiment, the evaluation image containing the above-mentioned brightness unevenness is corrected using a brightness correction coefficient to eliminate the brightness unevenness.

[0045] Returning to Figure 5, the CPU 140 then processes the corrected evaluation image (S23). In this image processing, the CPU 140 recognizes whether or not the evaluation image contains a welded area. Figure 7 shows an example of an evaluation image that contains a welded area. Since the welded area is formed across the material coil 200 from left to right in the width direction, in the evaluation image of Figure 7, the black, river-like area that divides the material coil 200 in the x-axis (front-back) direction corresponds to the welded area. However, since the characteristics of the welded area in the evaluation image are not constant due to differences in welding conditions, it was extremely difficult to distinguish between welded and non-welded areas by image recognition based on known characteristics of the welded area in the evaluation image. Therefore, the characteristics of the welded area in the captured image are learned by machine learning from a large number of captured images that include welded areas, and based on the learning results, a neural network consisting of, for example, three layers is constructed, and image recognition is performed using that neural network. Although not shown in the example, image recognition uses machine learning to recognize welds from patterns of welds, scratches, and dirt present in the evaluation image. However, if there is uneven brightness in the evaluation image, it cannot be recognized correctly. Therefore, the brightness correction coefficient generation process described above removes only the uneven brightness, and then represents the welds, scratches, and dirt present on the material coil 200 as they are, enabling recognition of only the welds. The verification of recognition through correction of uneven brightness was conducted by each company based on at least 3 million images of material coil 200 with a total length of approximately 5000m, and the recognition of welds is sufficiently practical.

[0046] Next, the CPU 140 determines whether or not a welded area has been detected by image recognition (S24). If a welded area is detected (S24: YES), the CPU 140 outputs a rejection signal to the molding machine 50 via the communication IF 160 (S25), and then proceeds to S26. On the other hand, if no welded area is detected (S24: NO), the CPU 140 skips S25 and proceeds to S26.

[0047] In S26, the CPU 140 determines whether or not to terminate the weld detection process. The decision to terminate the weld detection process is based on factors such as the operator of the molding system 1 pressing a stop button (not shown) or the end of the coil wound around the material coil 200. If the decision in S26 is to continue the weld detection process (S26: NO), the CPU 140 returns the process to S21 and continues the process from S21. On the other hand, if the decision in S26 is to terminate the weld detection process (S26: YES), the CPU 140 terminates the weld detection process.

[0048] In this embodiment, the welding detection process, specifically processes S21 to S25, must be completed in, for example, 12.8 msec or less per run. This is because the product is manufactured at high speed. Therefore, it is necessary to speed up each of the processes S21 to S25 as much as possible, but speeding up at least one of S22 and S23 will ultimately speed up the processes S21 to S25 as a whole. Accordingly, in this embodiment, the process of S22 is accelerated. In other words, the brightness correction coefficient used in the correction process of S22 is calculated in advance by the brightness correction coefficient generation process and stored in memory 150, and is simply read and used. Furthermore, in the correction process of S22, as described above, only the brightness value of each pixel constituting the evaluation image is multiplied by the brightness correction coefficient.

[0049] Furthermore, in order to improve the generation accuracy by the brightness correction coefficient generation process and the detection accuracy by the weld detection process, it is preferable that the entire material coil 200 is captured in the captured image (including both the reference image and the evaluation image) without including any other components. However, because it is difficult to restrict the widthwise vibration of the material coil 200 due to vibrations during product production and high-speed transport using rail gaps, if the entire width of the material coil 200 is included in the shooting range, a part of the material coil 200 may fall outside the shooting range. In this case, since a part of the material coil 200 is not included in the captured image, there is a risk that the weld area cannot be correctly detected. In addition, there are cases where processing associated with the detection of the weld area must be performed, such as processing to distinguish between the area of ​​the material coil 200 and other areas from the captured image, and if this is addressed, the weld detection process will be delayed. Therefore, in this embodiment, the shooting range is widened so that the entire width of the material coil 200 is always included, and the area in which the material coil 200 is visible is cropped from the captured image obtained by shooting within that shooting range, and the above-mentioned weld detection process is performed on that area.

[0050] Figure 8 shows the procedure for cropping the captured image to extract the region containing the material coil 200. The captured image cropping process can be inserted and executed between S22 and S23 (see Figure 5) of the weld detection process. However, the captured image cropping process in Figure 8 is configured as an independent process and includes the process of capturing the material coil 200 (S31, which will be explained next). Therefore, when inserting and executing the captured image cropping process during the weld detection process, the process of capturing the material coil 200 has already been performed in S21, so the process of S31 is excluded and executed.

[0051] In Figure 8, the CPU 140 first captures a predetermined area including the material coil 200, which is the material to be molded, in the same manner as in S21 (S31). However, while S21 captures the entire width of the material coil 200, S31 captures a predetermined area including the entire width of the material coil 200. Therefore, if the captured image trimming process is to be inserted and executed during the weld detection process, it is necessary to change the capture process in S21 during the weld detection process to the capture process in S31.

[0052] Next, the CPU 140 averages the brightness value of each pixel in the captured image along the x-axis (S32). Assuming that the captured image Ixy has C pixels along the x-axis, the average brightness value Ly bar in the x-axis direction of each pixel is calculated by the following equation (5).

[0053]

number

[0054] Figure 9(a) shows an example of an image captured by the processing in S31 described above. Figure 9(b) shows the average luminance distribution, i.e., the luminance projection histogram in the x-axis direction, obtained by performing the processing in S32 described above on the image captured in Figure 9(a). In Figure 9(b), the vertical axis represents the average luminance value Ly in the x-axis direction, and the horizontal axis represents the y-axis, i.e., the width direction of the material coil 200. As can be seen from the image captured in Figure 9(a), in an image obtained by capturing a predetermined area including the material coil 200, pixels outside the area of ​​the material coil 200 are black. In other words, the pixel values ​​outside the area of ​​the material coil 200 are values ​​very close to "0". Since the material coil 200 is captured from end to end in the x-axis direction in the image, the average luminance distribution is obtained by averaging each luminance value in the x-axis direction, i.e., the luminance projection histogram in the x-axis direction.

[0055] Returning to Figure 8, the CPU 140 then determines a trimming threshold based on the luminance values ​​averaged along the x-axis, i.e., the luminance projection histogram in the x-axis direction (S33). In the average luminance distribution in Figure 9(b), the threshold Th shows an example of a trimming threshold determined by the processing in S33.

[0056] Next, the CPU 140 trims the pixel range with brightness values ​​exceeding the trimming threshold from the captured image (S34), and then terminates the captured image trimming process. When the captured image trimming process is performed on the captured image in Figure 9(a), the area of ​​the material coil 200 is trimmed from the captured image, so the captured image after trimming becomes an image that shows only the material coil 200, like the captured image in Figure 6(a), compared to the captured image in Figure 9(a). Therefore, if the weld detection process is performed on this image, the accuracy of detecting the welded area is improved. In addition, the weld detection process does not require processing that is associated with detecting the welded area, such as processing to distinguish the area of ​​the material coil 200 from other areas in the captured image, so the weld detection process can be shortened.

[0057] Furthermore, if the captured image trimming process is performed independently rather than within the weld detection process, it is preferable to insert the process in S22 (Figure 5) between S31 and S32. This is because, as mentioned above, the captured image has uneven brightness, which may prevent the determination of an appropriate trimming threshold. Alternatively, the captured image trimming process can also be performed in the brightness correction coefficient generation process. In the captured image trimming process, it can be inserted and executed between S14 and S15 (see Figure 4).

[0058] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are possible without departing from the spirit of the invention. (1) In the above embodiment, a material coil 200 was used as the material to be inspected, and a welded area was used as the abnormal area. However, the present invention is not limited to these, and may be applied to sorting for the presence or absence of scratches on objects being continuously transported, or to sorting for the external shape of such objects. (2) In the above embodiment, a PC100 was used as the device for controlling the molding system 1, but the device is not limited to this and may be a portable device such as a smartphone or tablet. [Explanation of Symbols]

[0059] 1... Molding system, 10... Powered material stand, 20... Inspection device, 21... Camera, 22... Lighting fixture, 30... Material straightening roller, 40... Material feeder, 50... Molding machine, 100... PC, 130... Input / Output IF, 140... CPU, 150... Memory, 160... Communication IF, 200... Material coil, 201... Welding area.

Claims

1. A camera to photograph the material being inspected, A lighting fixture that illuminates the shooting range of the camera with a second optical axis in a direction different from the first optical axis of the camera, A control processing device that uses the captured image obtained by the aforementioned camera as an evaluation image, performs image processing on the evaluation image, and determines whether or not the material to be inspected contains an abnormal part, An inspection device having, The control processing device is A first averaging process is performed to generate a luminance projection histogram by taking a photograph of the material to be inspected that does not contain the aforementioned abnormal parts with the camera, using one or more captured images as reference images, and averaging the luminance values ​​of each pixel in the reference images over an axis orthogonal to the axis generated by projecting the second optical axis onto the image plane of the reference image, The luminance projection histogram generated by the first averaging process is approximated by an approximation function composed of an n-th degree polynomial (where n is an integer value of 3 or more), and then the reciprocal of the approximated luminance projection histogram is calculated, or the reciprocal of the luminance projection histogram is calculated and then approximated by the approximation function to obtain a luminance correction value. A correction process that corrects the evaluation image acquired by the camera using the brightness correction value obtained by the calculation process, Execute, In the aforementioned image processing, image processing is performed on the evaluation image after correction by the correction process to determine whether or not the material being inspected contains any abnormal parts. Inspection device.

2. The camera captures a predetermined area including the material to be inspected. The control processing device further, A second averaging process is performed to average the brightness values ​​of each pixel in the captured image, including the reference image and the evaluation image, obtained by capturing the predetermined range with the camera, over a direction along the edge of the material under inspection. A determination process for determining a trimming threshold based on the averaged brightness value obtained by the second averaging process, A trimming process is performed to trim from the captured image the pixel range having a brightness value exceeding the trimming threshold determined by the aforementioned determination process. Execute, The cropped image obtained by the aforementioned cropping process is passed to the first averaging process and the correction process. The inspection apparatus according to claim 1.

3. An inspection method comprising a camera for photographing the material to be inspected, and an illumination device for illuminating the camera's shooting range with a second optical axis in a direction different from the camera's first optical axis, wherein the image obtained by the camera is used as an evaluation image, and image processing is performed on the evaluation image to determine whether or not the material to be inspected contains any abnormal parts, A first averaging step involves taking a photograph of the material to be inspected, which does not contain the aforementioned abnormal area, with the camera and obtaining one or more captured images, which are used as reference images. The luminance values ​​of each pixel in the reference images are averaged over an axis orthogonal to the axis generated by projecting the second optical axis onto the image plane of the reference image to generate a luminance projection histogram. A calculation step is to approximate the luminance projection histogram generated by the first averaging step with an approximation function composed of an n-th degree polynomial (where n is an integer value of 3 or more), then calculate the reciprocal of the approximated luminance projection histogram, or calculate the reciprocal of the luminance projection histogram and then approximate it with the approximation function to obtain a luminance correction value. A correction step in which an evaluation image acquired by shooting with the aforementioned camera is corrected by a brightness correction value acquired in the calculation step, Includes, In the aforementioned image processing, image processing is performed on the evaluation image after correction by the correction step to determine whether or not the material under inspection contains any abnormal parts. Testing method.

Images