Punching sheet overlying detection system and method based on visual monitoring

Abstract

The invention discloses a punching sheet overlying detection system and method based on visual monitoring, and relates to the technical field of image processing. The system comprises a mechanical transmission module, an image acquisition module, a transmission control module, an image segmentation module and an image analysis module, and the transmission control module controls a transfer table to stop transmission and controls the image acquisition module to start image acquisition according to pressure data of a detection position and transmission time; controlling the transfer table to recover conveying according to the image acquisition duration; the image segmentation module carries out image fusion on each punching sheet surface image and each surface area image with the same secondary number according to an image segmentation rule; and comparing each fusion region image with the corresponding standard secondary image to determine the similarity of each fusion region image so as to determine whether the punching sheet corresponding to the folder is qualified or not. According to the method, the surface images of the lamination at multiple light source angles are obtained, the surface information of the lamination is comprehensively captured, and possible defect details are prevented from being missed.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a punching sheet stacking detection system and method based on visual monitoring. Background Art

[0002] In modern industrial production, punching sheets are widely used in many electrical equipment fields such as motors and transformers. The performance and reliability of these equipment depend to a large extent on the quality of the punching sheets. With the development of science and technology, industrial equipment is moving towards high performance and high precision, which puts forward more stringent requirements on the quality of punching sheets. Minor defects such as missing corners, deformation, oil stains or rust may lead to reduced equipment performance, shortened life or even failure. Therefore, a high-precision detection system is needed to ensure the quality of punching sheets.

[0003] The prior art provides a visual inspection device for stamped terminals, including a transmission module, a shooting module and an image processing module; the transmission module is used to transmit the stamped terminals to be quality inspected to the bottom of the shooting module; the shooting module is used to shoot the stamped terminals to obtain the appearance image of the stamped terminals; the image processing module is used to adjust the brightness of the appearance image, and then perform image recognition processing to determine whether the stamped terminal meets the preset quality requirements; the brightness adjustment processing of the appearance image includes: classifying the pixels in the appearance image, dividing the pixels into one type of pixels and two types of pixels; using a preset first processing function to adjust the brightness of the first type of pixels, and using a preset second processing function to adjust the brightness of the second type of pixels to obtain a brightness adjusted image. The above scheme effectively improves the accuracy of the brightness processing results, thereby improving the accuracy of the quality inspection results. It can be seen that the invention does not consider the problem of low detection accuracy caused by the inability to fully capture the surface information of the laminated sheets due to the single light source shooting the image. Summary of the invention

[0004] The purpose of this application is to provide a punching sheet stacking detection system and method based on visual monitoring, so as to solve the problem of low detection accuracy in the prior art due to the inability to fully capture the surface information of the stacked sheets due to the single light source shooting the image.

[0005] On the one hand, the embodiment of the present application provides a punching sheet stacking detection system based on visual monitoring, comprising:

[0006] The mechanical transmission module includes a transfer platform and a mechanical arm, wherein the mechanical arm is used to carry the punching sheet to the transmission start end of the transfer platform, and the transfer platform is used to transmit the punching sheet from the transmission start end to the detection position;

[0007] The image acquisition module includes a fill light unit for providing a plurality of shooting light sources at different angles, an image acquisition unit for acquiring an image of a punched film surface, and a shooting control unit for controlling the fill light unit to sequentially turn on the shooting light sources at various angles and controlling the image acquisition unit to perform image acquisition and pack a plurality of punched film surface images of a single punched film into a folder and generate a primary number of each punched film surface image according to the image acquisition sequence;

[0008] A transmission control module, which is connected to the mechanical transmission module and the image acquisition module respectively, and is used to control the transfer platform to stop transmission and control the image acquisition module to start image acquisition according to the pressure data of the detection position and the transmission time, and to control the transfer platform to resume transmission according to the image acquisition time;

[0009] An image segmentation module, which is connected to the image acquisition module, is used to divide each punched sheet surface image into a plurality of surface region images according to an image cutting rule and generate corresponding secondary numbers, and to fuse the surface region images with the same secondary number in the folder to form a corresponding fused region image;

[0010] An image analysis module, which is connected to the image segmentation module, compares each of the fusion area images in a single folder with the corresponding standard secondary image to determine the similarity of each fusion area image, and determines whether the punched film corresponding to the folder is qualified according to the similarity of each fusion area image;

[0011] Among them, in a single image acquisition, the number of punched film surface images generated is equal to the number of shooting light sources turned on, the surface area image includes a center area image and an edge area image, and the standard secondary image is a surface area image of the punched film standard part at the same position.

[0012] According to the punching sheet stacking detection system based on visual monitoring in an embodiment of the present application, the detection position is located on the base of the transfer platform, and a pressure sensor for determining the pressure data of the transfer platform on the detection position is arranged between the detection position and the transfer platform.

[0013] According to the punching sheet stacking detection system based on visual monitoring in the embodiment of the present application, the transmission control module determines whether the punching sheet reaches the detection position according to the pressure data and transmission time of the detection position, wherein:

[0014] If the pressure data is within the reference pressure range and the transmission time is within the preset time range, it is determined that the punching sheet has reached the detection position;

[0015] If the pressure data is not within the reference pressure range and / or the transmission time is not within the preset time range, it is determined that the punching sheet has not reached the detection position.

[0016] According to the punching sheet stacking detection system based on visual monitoring in the embodiment of the present application, the conveying control module controls the transfer table to stop conveying and controls the image acquisition module to start image acquisition according to the determination result that the punching sheet has arrived at the detection position;

[0017] Wherein, controlling the image acquisition module to start image acquisition includes:

[0018] The shooting control unit controls the fill light unit to sequentially turn on the shooting light sources at various angles and controls the image acquisition unit to perform image acquisition when turning on the shooting light sources at various angles.

[0019] According to the punching sheet stacking detection system based on visual monitoring in the embodiment of the present application, the conveying control module determines the end of the image acquisition of a single punching sheet according to the image acquisition time of the single punching sheet and controls the transfer table to resume conveying, wherein:

[0020] If the image acquisition time of a single punched sheet is equal to the preset time, it is determined that the image acquisition of the punched sheet is completed and the transfer table is controlled to resume transmission.

[0021] According to the punching sheet stacking detection system based on visual monitoring in the embodiment of the present application, the image segmentation module pre-stores an image cutting rule, and the image cutting rule includes determining to divide each punching sheet surface image into an edge image and a center image according to a reference area ratio, and determining a first reference number and a second reference number according to the detection accuracy to respectively divide the edge image and the center image into a first reference number of edge area images and a second reference number of center area images;

[0022] The central image and the corresponding punching surface image are similar images, and the geometric center of the central image is the geometric center of the corresponding punching surface image.

[0023] According to the punching sheet overlap detection system based on visual monitoring in the embodiment of the present application, the reference area ratio is the area ratio of the center image to the punching sheet surface image.

[0024] According to the punching sheet overlay detection system based on visual monitoring in the embodiment of the present application, a machine learning model is preset in the image analysis module to determine the similarity of the images of each fusion area.

[0025] According to the punching sheet stacking detection system based on visual monitoring in the embodiment of the present application, the image analysis module determines whether the punching sheet corresponding to the folder is qualified according to the similarity of the images of each fusion area, including:

[0026] Whether the corresponding punched film has missing corners is determined based on the similarity between the fused area images corresponding to each edge area image, and whether the corresponding punched film has defects is determined based on the similarity between the fused area images corresponding to each center area image.

[0027] On the other hand, the embodiment of the present application further proposes a punching sheet stacking detection method based on visual monitoring, comprising:

[0028] Step S1, obtaining pressure data and transmission time data of the detection position in real time;

[0029] Step S2, determining whether the punching sheet reaches the detection position according to the pressure data and the transmission time at the detection position;

[0030] Step S3, according to the judgment result of reaching the detection position, controlling the transfer platform to stop conveying and controlling the image acquisition module to start image acquisition, and controlling the mechanical arm to carry the punched sheet to the conveying start end of the transfer platform;

[0031] Step S4, controlling the transfer table to resume conveyance according to the image acquisition time of the punched film.

[0032] In summary, the beneficial technical effects of the present application are as follows: the punching sheet stacking detection system based on visual monitoring provided by the present application provides an efficient and accurate solution for punching sheet detection through the coordinated work of the mechanical transmission module, the image acquisition module, the transmission control module, the image segmentation module and the image analysis module; among them, the transfer table and the robotic arm of the mechanical transmission module cooperate to ensure that the punching sheet can accurately reach the detection position, and the image acquisition module utilizes a multi-angle fill light unit and a high-resolution camera to comprehensively obtain the surface image of the punching sheet under different light source angles of the stacked sheets and reasonably number each surface image of the punching sheet; the transmission control module accurately detects the punching sheet based on the pressure at the detection position and the transmission time. The coordination of controlling the transfer table and image acquisition ensures the integrity and accuracy of image acquisition; the image segmentation module divides the surface image into different areas according to the pre-stored image cutting rules, and fuses the surface area images at different positions using a suitable fusion method to highlight the oil and rust characteristics of the center area image and the missing corners and deformation characteristics of the edge area image; the image analysis module compares the image of each fused area with the corresponding standard secondary image to determine whether the punching is qualified based on the similarity of the comparison results. The entire system effectively ensures the accuracy of detection and improves detection efficiency in batch production detection based on the standard secondary image determined by the previous standard parts;

[0033] The transmission control module combines the pressure data and transmission time of the detection position to accurately determine whether the punching sheet has reached the detection position, wherein the pressure data is based on the reference pressure range determined by the quality of the standard parts of the punching sheet, and the transmission time is based on the preset time range determined by the spacing between adjacent punching sheets and the transmission speed. This dual-factor judgment mechanism ensures the accuracy of the judgment; in addition, in actual operation, the appropriate transmission speed set considering the high-precision detection requirements of the punching sheets, as well as the setting of the time error as small as possible, are helpful to more accurately control the stop timing of the transfer table to ensure that the punching sheet is accurately within the shooting range of the image acquisition unit; once it is determined that the punching sheet has reached the detection position, the transmission control module can effectively control the transfer table to stop transmission, and accurately command the shooting control unit to turn on the light sources of the fill light unit at various angles, and at the same time let the image acquisition unit perform image acquisition when each light source is turned on, so that the entire image acquisition process is closely coordinated with the transmission of the punching sheet in an orderly manner, greatly improving the efficiency of the detection system and the reliability of the detection results;

[0034] The image segmentation module stores image cutting rules that determine the division of the surface image of the punch into edge image and center image by reference area ratio, and determines the first reference number and the second reference number respectively according to the detection accuracy, thereby dividing the edge image and the center image into the corresponding number of regional images; this design can reasonably determine the number of pixel points that make up the surface regional image according to the detection accuracy (high accuracy, medium accuracy, low accuracy), avoiding unnecessary calculation and resource waste, and at the same time, by reasonably setting the reference area ratio, it ensures the accuracy of using the center part of the punch to detect oil stains and rust and using the edge part to detect missing corners or deformation, making image segmentation more targeted and scientific, effectively improving the efficiency and quality of subsequent image analysis, and laying a solid foundation for the entire detection system to accurately judge whether the punch is qualified;

[0035] The machine learning model preset inside the image analysis module can accurately determine the similarity of each fused area image, and judge whether the punched film has missing corners or defects through the similarity between the fused area images corresponding to each edge area image and the central area image and their corresponding standard secondary images; and setting the preset similarity according to different detection accuracies enables the image analysis module to efficiently and accurately evaluate the quality of the punched film and effectively screen out defective products, providing reliable guarantee for ensuring the quality of the punched film and greatly improving the qualified rate and quality stability of the products in the entire production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a connection diagram of a punching sheet stacking detection system based on visual monitoring according to an embodiment of the present application;

[0037] Figure 2 It is a schematic diagram of the composition of a single folder in an embodiment of the present application;

[0038] Figure 3It is a step diagram of a punching sheet stacking detection method based on visual monitoring in an embodiment of the present application. DETAILED DESCRIPTION

[0039] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0040] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0041] The term “plurality” used in this application refers to two or more (including two).

[0042] On the one hand, the following combination Figure 1 to Figure 2 The punching sheet stacking detection system based on visual monitoring in an embodiment of the present application is described in detail.

[0043] See attached Figure 1 and attached Figure 2 , which are respectively a connection diagram of a punching sheet stacking detection system based on visual monitoring according to an embodiment of the present application and a schematic diagram of the composition of a single folder according to an embodiment of the present application.

[0044] The present application provides a punching sheet stacking detection system based on visual monitoring, comprising:

[0045] The mechanical transmission module includes a transfer platform and a mechanical arm, wherein the mechanical arm is used to carry the punch to the conveying start end of the transfer platform, and the transfer platform is used to convey the punch from the conveying start end to the detection position; it can be understood that the mechanical arm carries the punch to the start end of the transfer platform and stably conveys it to the detection position through the transfer platform, thereby ensuring the orderly transportation of the punch in the detection process, improving the efficiency of the overall detection process, and avoiding problems such as position deviation that may be caused by manual handling;

[0046] The image acquisition module includes a fill light unit for providing a plurality of shooting light sources at different angles, an image acquisition unit for acquiring an image of a punched film surface, and a shooting control unit for controlling the fill light unit to sequentially turn on the shooting light sources at various angles and controlling the image acquisition unit to perform at least one image acquisition (usually one image is sufficient for one position) and to package a plurality of punched film surface images of a single punched film into a folder and generate a primary number for each punched film surface image according to the image acquisition sequence;

[0047] It can be understood that the fill light unit can be a fill light device capable of providing multi-angle light sources in the prior art, and the image acquisition unit can be a camera with a shooting function in the prior art (in implementation, a high-definition camera with high resolution is preferably used); in addition, the image acquisition unit is usually fixed at the center of the fill light unit, and the plane formed by the camera lens and each light source of the image acquisition unit should be parallel to the transfer platform, and the straight line formed by the image acquisition unit and the detection position is perpendicular to the transfer platform;

[0048] In implementation, the fill light unit is fixedly connected to both sides of the base of the transfer platform through two fixed support rods, and the plane formed by the two fixed support rods is perpendicular to the plane of the base;

[0049] Attached Figure 2 It is shown that there are n light sources at different angles and a single film surface image is cut into m surface area images according to the image cutting rules, and also includes m fusion area images, which are respectively recorded as m1...mm. It can be understood that for any numbered film (any film will have its own unique identification code when it is prepared, and the number here is the corresponding identification code), the film surface images of n light sources at different angles are collected and numbered according to the order of collection (that is, if the fill light unit has n light sources at different angles, n film surface images will be collected and numbered as 1, 2, 3,..., n), and the film surface images of these n light sources at different angles are enclosed in a folder (the name of this folder is the number of the film); therefore, in one implementation, the fill light unit has 3 light sources at different angles, and a film is numbered AA, then the folder named AA should contain film surface images named 1, 2, and 3 respectively;

[0050] It is understandable that in practice, the surface image acquisition process at a single light source angle may ignore some surface defects, that is, the application of light will affect the analysis of the image surface. Therefore, the fill light unit provides multiple shooting light sources at different angles to cooperate with the image acquisition unit, which can obtain surface images of multiple light source angles of the punched film, and comprehensively capture the surface information of the stacked film to avoid missing possible defect details; the shooting control unit numbers the acquired images in sequence and packages them into folders, and each punched film has its unique identifier, which is convenient for subsequent image processing and analysis and avoids confusion of image data; the image acquisition unit is fixed at the center of the fill light unit, and the plane formed by the camera lens and the light sources is parallel to the transfer table and perpendicular to the transfer table with the straight line formed by the detection position. At the same time, the fill light unit is connected to the transfer table in a specific way, which ensures the stability and consistency of the acquired image and reduces the image error caused by unstable equipment position;

[0051] A transmission control module, which is connected to the mechanical transmission module and the image acquisition module respectively, and is used to control the transfer platform to stop transmission and control the image acquisition module to start image acquisition according to the pressure data of the detection position and the transmission time, and to control the transfer platform to resume transmission according to the image acquisition time;

[0052] It can be understood that the transmission control module can accurately control the transfer table to stop and the image acquisition to start according to the pressure data and transmission time of the detection position, and can also control the transfer table to resume transmission according to the image acquisition time, thereby realizing the precise coordination of mechanical transmission and image acquisition, ensuring that each punching sheet can be accurately imaged at the appropriate position and time, thereby improving the accuracy and efficiency of detection;

[0053] An image segmentation module is connected to the image acquisition module and is used to receive each punched sheet folder sent by the shooting control unit, divide each punched sheet surface image in the folder into a plurality of surface region images according to the image segmentation rule and generate corresponding secondary numbers, and fuse each surface region image with the same secondary number in the folder to form a corresponding fused region image;

[0054] It can be understood that any punched sheet surface image determined according to the image cutting rule is divided into the same number of surface region images, and each surface region image with the same secondary number should be at the same position of the punched sheet surface image;

[0055] It can be understood that the name of each surface area image is a combination of a primary number and a secondary number, and the combination can be in any manner as long as the primary number part and the secondary number part can be distinguished; in one implementation, the primary number and the secondary number of a surface area image are 1 and 02 respectively, then the name of the surface area image can be 102, 1.02, 1-02, 10002, etc.;

[0056] In the above implementation, there are three film surface images named 1, 2, and 3 in the folder named AA, and the image cutting rule has determined how to divide the film surface image into m regional surface images and determine that the regional surface images at each position are numbered one of 1 to m; the film surface image named 1 can be divided into m regional surface images and named 1-1, 1-2, 1-3, ..., 1-m respectively, the film surface image named 2 can be divided into m regional surface images and named 2-1, 2-2, 2-3, ..., 2-m respectively, and the film surface image named 3 can be divided into m regional surface images and named and 3-1, 3-2, 3-3, ..., 3-m; at this time, there should be 3 film surface images and 3m regional surface images in the folder named AA, and these 3m regional surface images are 1-1, 1-2, 1-3, ..., 1-m, 2-1, 2-2, 2-3, ..., 2-m, 3-1, 3-2, 3-3, ..., 3-m; then for the three regional surface images 1-1, 2-1 and 3-1 with the secondary number 1, they are in the same position in their corresponding film surface images. If 1-1 is in the upper right corner of the film surface image named 1, then 2-1 and 3-1 are also in the upper right corners of the film surface images named 2 and 3 respectively;

[0057] In implementation, when the surface area images with the same secondary number in the folder are fused to form a corresponding fused area image, different fusion methods are determined according to the surface area images, including:

[0058] (1) The principal component analysis (PCA) fusion method is used to fuse the central area images with the same secondary number to form a fused area image with the secondary number. It can be understood that PCA can help screen out the most important information for fusion, highlight the main features while reducing data redundancy, and highlight the main feature areas of these defects when detecting oil stains and rust on laminated sheets.

[0059] (2) Using a fusion method based on multi-resolution analysis (such as wavelet transform) to fuse the edge area images with the same secondary number to form a fused area image with the secondary number; it can be understood that wavelet transform can extract the most useful edge information from photos at different angles and fuse them together to obtain a clearer and more accurate edge image. When detecting whether the edge of the stacked film is missing or deformed, wavelet transform fusion can better highlight the edge details;

[0060] It can be understood that after the punching surface image is divided into several surface area images according to the image cutting rules, the principal component analysis (PCA) fusion method is used for the central area image to highlight the main characteristic areas of defects such as oil stains and rust; the fusion method based on multi-resolution analysis (such as wavelet transform) is used for the edge area image to better highlight the edge details, such as missing corners or deformation. This targeted processing helps to more accurately detect different types of defects;

[0061] An image analysis module, which is connected to the image segmentation module, compares each of the fusion area images in a single folder with the corresponding standard secondary image to determine the similarity of each fusion area image, and determines whether the punched film corresponding to the folder is qualified according to the similarity of each fusion area image;

[0062] It can be understood that the standard secondary image corresponding to the fused region image has the same secondary number as the fused region image;

[0063] It can be understood that the comparison method of comparing the fusion area image with the corresponding standard secondary image to determine the similarity and then judge whether the punched film is qualified can effectively ensure product quality in batch production inspection, ensure that unqualified punched films are accurately screened out, and improve the quality control level of the entire production process; in addition, before batch production inspection, the standard secondary image is determined by preparing punched film standard parts, which provides a reliable basis for subsequent comparative analysis, making the inspection results more authoritative and accurate;

[0064] Among them, in a single image acquisition, the number of punched film surface images generated is equal to the number of shooting light sources turned on, the surface area image includes a center area image and an edge area image, and the standard secondary image is a surface area image of the punched film standard part at the same position.

[0065] It is understandable that, before mass production testing, a number of punching standard parts are prepared in the experiment, and a standard secondary image is determined based on each punching standard part.

[0066] Specifically, the detection position is located on a base of the transfer platform, and a pressure sensor for determining pressure data of the transfer platform on the detection position is provided between the detection position and the transfer platform.

[0067] In implementation, the pressure sensor can be any high-precision pressure sensor in the prior art; it can be understood that high-precision pressure measurement can avoid the transfer table from stopping too early or too late due to pressure data errors, thereby ensuring that each punch can be imaged at the optimal position, improving the accuracy and reliability of the detection, and ensuring the smooth progress of the entire detection process.

[0068] Specifically, the transmission control module determines whether the punching sheet reaches the detection position according to the pressure data and transmission time of the detection position, wherein:

[0069] If the pressure data is within the reference pressure range and the transmission time is within the preset time range, it is determined that the punching sheet has reached the detection position;

[0070] If the pressure data is not within the reference pressure range and / or the transmission time is not within the preset time range, it is determined that the punching sheet has not reached the detection position.

[0071] It can be understood that the reference pressure range is determined according to the mass of the punching standard parts, and the average mass m and mass standard deviation m1 of several punching standard parts prepared in the experiment are determined. The reference pressure range is (m-m1)×g~(m+m1)×g (g is the acceleration of gravity, which is about 9.8N / kg); in the implementation, the unit of mass is kg, and the unit of reference pressure is N;

[0072] It can be understood that when a punch is delivered to the detection position, a new punch is placed at the beginning of the delivery, and then when the next punch is delivered to the detection position, a new punch is placed at the beginning of the delivery. When the delivery speed of the transfer table remains unchanged, the distance between any two adjacent punches should be equal;

[0073] It can be understood that the preset time range is determined according to the spacing between two adjacent punches on the transfer table and the transmission speed, and the standard preset time = the spacing between two adjacent punches on the transfer table ÷ the transmission speed; in practice, the accuracy requirement of the punches is relatively high, so the transmission speed is not very fast, generally 1m / min~5m / min, and is mostly set to 2m / min. This speed range not only meets the basic requirements of production efficiency, but also ensures the stability of the punches during the transmission process, avoids image acquisition blur or other problems caused by too fast speed, and lays the foundation for subsequent accurate detection; in practice, the preset time range ∈ [standard preset time-time error, standard preset time+time error], generally the time error is ≤1s, in order to ensure that the punch is within the shooting range of the image acquisition unit when the transfer table stops, the time error should be as small as possible, preferably set to 0.5s, introduce the time error in the setting of the preset time range and try to control it to a smaller value (generally ≤1s, preferably 0.5s), which can more accurately control the stop time of the transfer table, ensure that the punch is within the shooting range of the image acquisition unit when the transfer table stops, and further improve the quality of image acquisition and the reliability of the detection results;

[0074] It can be understood that the transmission control module uses two factors, pressure data and transmission time, to determine whether the punch has reached the detection position, and comprehensively considers the weight of the punch and the time factors in the transmission process. This method is more accurate than a single factor judgment and reduces the possibility of misjudgment; the reference pressure range is determined according to the mass of the punch standard parts, and the range is set by calculating the average mass and quality standard deviation of the standard parts, so that the pressure judgment is closely related to the physical properties of the actual punch, ensuring that punches of different batches but that meet the standards can be accurately identified, and effectively avoiding position judgment errors caused by individual quality differences of the punches; the preset time range is determined according to the spacing between adjacent punches on the transfer table and the transmission speed. This method fully considers the dynamic factors in the transmission process. In actual production, since the spacing between punches and the transmission speed are relatively stable, the position of the punch can be accurately judged by reasonably setting the preset time range to ensure the timeliness and accuracy of the detection.

[0075] Specifically, the conveying control module controls the transfer platform to stop conveying and controls the image acquisition module to start image acquisition according to the determination result that the punching sheet arrives at the detection position;

[0076] Wherein, controlling the image acquisition module to start image acquisition includes:

[0077] The shooting control unit controls the fill light unit to sequentially turn on the shooting light sources at various angles and controls the image acquisition unit to perform at least one image acquisition when turning on the shooting light sources at various angles.

[0078] It can be understood that when it is determined that the film has arrived at the detection position, the transmission control module promptly controls the transfer table to stop transmission and simultaneously starts the image acquisition module to start image acquisition. This rapid coordination capability ensures the seamless connection between image acquisition and film transmission, and avoids problems such as omission or incomplete image acquisition caused by time delays or asynchronous operations. In the image acquisition process, the shooting control unit, under the command of the transmission control module, orderly controls the fill light unit to turn on the shooting light sources at various angles in turn, and allows the image acquisition unit to perform at least one image acquisition when each light source is turned on. This standardized operating procedure ensures that the image information of the film is fully acquired from different angles, providing a rich and accurate data source for subsequent image analysis.

[0079] Specifically, the conveying control module determines the end of image acquisition of a single punched sheet according to the image acquisition time of the single punched sheet and controls the transfer table to resume conveying, wherein:

[0080] If the image acquisition time of a single punched sheet is equal to the preset time, it is determined that the image acquisition of the punched sheet is completed and the transfer table is controlled to resume transmission.

[0081] It is understandable that the fill light unit can have 3 to 10 light sources at different angles. Setting too many light source angles will lead to data redundancy and greatly increase the time and resource waste of analysis and processing. Therefore, usually setting 4 light sources at different angles can achieve accurate analysis of defects on the surface of the punched film; determine the average time of turning on and shooting the light sources at each angle in advance experiments, and determine the product of the number of light sources at different angles and the average time as the preset duration. In one implementation, the preset duration is 4 times the average time.

[0082] It can be understood that the transmission control module determines whether the acquisition is completed based on the image acquisition time of a single film and controls the transfer table to resume transmission, and performs corresponding operations when the image acquisition time is equal to the preset time. This control method reasonably sets the number of light source angles of the fill light unit, and determines the average time of turning on and shooting the light sources at each angle in the experiment in advance to calculate the preset time, thereby accurately coordinating the image acquisition and transfer table transmission processes to ensure efficient and stable operation of the detection system, which not only ensures the integrity of the film image acquisition, but also avoids unnecessary waste of time, effectively improving the efficiency and reliability of the entire detection process.

[0083] Specifically, the image segmentation module pre-stores an image cutting rule, the image cutting rule comprising determining to divide each punching sheet surface image into an edge image and a center image according to a reference area ratio, and determining a first reference number and a second reference number according to the detection accuracy to respectively divide the edge image and the center image into a first reference number of edge area images and a second reference number of center area images;

[0084] The central image and the corresponding punching surface image are similar images, and the geometric center of the central image is the geometric center of the corresponding punching surface image.

[0085] It can be understood that the area of ​​the center image is much larger than the area of ​​the edge area, so the first reference number is much smaller than the second reference number; in implementation, each surface area image detected with high precision is composed of 1 to 4 adjacent pixels, each surface area image detected with medium precision is composed of 1 to 10 adjacent pixels, and each surface area image detected with low precision is composed of 1 to 20 adjacent pixels;

[0086] In implementation, the second reference number and the first reference number are determined according to the number of pixels of the center image and the edge image; for example, in high-precision detection, the pixel of the center image is a1, and the pixel of the edge image is a2: if a1÷4 is a positive integer, the center area image consists of 4 pixels; if a1÷4 is not a positive integer, it is determined whether a1÷3 is a positive integer; if a1÷3 is a positive integer, the center area image consists of 3 pixels; if a1÷3 is not a positive integer, it is determined whether a1÷2 is a positive integer; if a1÷2 is a positive integer, the center area image consists of 2 pixels; if a1÷2 is not a positive integer, the center area image consists of 1 pixel; it can be understood that the method for determining the number of pixels constituting a single edge area image is the same;

[0087] It can be understood that when determining the number of pixel points constituting the surface area image, the maximum number within each precision detection range is used as the starting point.

[0088] It can be understood that the first benchmark number and the second benchmark number are determined respectively according to the detection accuracy to further divide the edge image and the center image into a suitable number of regional images. This method matches the image segmentation with the detection accuracy, and can better adapt to and meet the needs of detection scenarios with different accuracy requirements, avoiding detection errors caused by unreasonable segmentation. Through reasonable image cutting rules, it avoids the increase in computational complexity and waste of resources caused by unnecessary fine segmentation or overly coarse segmentation, so that the detection system can maintain efficient operation while meeting the accuracy requirements.

[0089] Specifically, the reference area ratio is the area ratio of the center image to the image on the punching surface.

[0090] It can be understood that the reference area ratio is less than 1. In this application, the chipping or deformation is determined based on the edge of the punching sheet, and the oil stains and rust are determined based on the center of the punching sheet. Therefore, the area of ​​the center should be as large as possible, so the reference area ratio is ≥0.8 and the higher the detection accuracy, the larger the value of the reference area ratio; in implementation, for high-precision detection (detection accuracy ≤0.5mm), the reference area ratio is preferably 0.9, for medium-precision detection (0.5mm<detection accuracy ≤2mm), the reference area ratio is preferably 0.85, and for low-precision detection (2mm<detection accuracy), the reference area ratio is preferably 0.8.

[0091] It can be understood that the image segmentation module uses the reference area ratio to divide the edge image and the center image. Since the chipping or deformation is determined based on the edge part and the oil stain and rust are determined based on the center part when detecting punching, a reasonable reference area ratio can ensure that the area of ​​the center part is large enough, which is conducive to the accurate detection of different types of defects and improves the targeted nature of defect detection.

[0092] Specifically, a machine learning model is preset in the image analysis module to determine the similarity between each fusion area image and the corresponding standard secondary image.

[0093] It is understandable that the machine learning model has a strong data analysis capability and can quickly and accurately process image data, extract features and calculate similarity, which is more efficient and accurate than traditional methods. By analyzing the similarity of the fused area image corresponding to the edge area image and the center area image, the situation of different parts of the punched film is fully considered. It can detect the missing corner problem of the edge part and the defects of the center part (such as oil stains and rust), realizing a comprehensive evaluation of the quality of the punched film.

[0094] In implementation, the machine learning model is retrained once a week to ensure the accuracy of its judgments.

[0095] Specifically, the image analysis module determines whether the punched film corresponding to the folder is qualified according to the similarity of the images in each fusion area, including:

[0096] Whether the corresponding punched film has missing corners is determined based on the similarity between the fused area images corresponding to each edge area image, and whether the corresponding punched film has defects is determined based on the similarity between the fused area images corresponding to each center area image.

[0097] In implementation, if the similarity of the fused area image corresponding to any edge area image is less than the preset similarity, the corresponding punched sheet is determined to have a missing corner; if the similarity of the fused area image corresponding to any center area image is less than the preset similarity, the corresponding punched sheet is determined to have a defect (oil stain / rust);

[0098] It can be understood that the preset similarity is greater than 70%, and the higher the detection accuracy, the greater the preset similarity. Usually: the preset similarity of high-precision detection is ≥85%, the preset similarity of medium-precision detection is ≥80%, and the preset similarity of low-precision detection is ≥75%.

[0099] It is understandable that the workflow and judgment criteria of the entire image analysis module ensure the stability of the quality assessment of the punched sheet; different batches of production products can be tested according to the same rules, so that the product quality is not overly affected by factors such as production batches and environment, ensuring the consistency and stability of product quality, which is conducive to maintaining the company's production reputation and market competitiveness.

[0100] On the other hand, the following combination Figure 3 The punching sheet overlap detection method based on visual monitoring in an embodiment of the present application is described in detail.

[0101] See attached Figure 3, which is a step diagram of a punching sheet overlap detection method based on visual monitoring in an embodiment of the present application. The present application also provides a punching sheet overlap detection method based on visual monitoring, comprising:

[0102] Step S1, obtaining pressure data and transmission time data of the detection position in real time;

[0103] Step S2, determining whether the punching sheet reaches the detection position according to the pressure data and the transmission time at the detection position;

[0104] Step S3, according to the judgment result of reaching the detection position, controlling the transfer platform to stop conveying and controlling the image acquisition module to start image acquisition, and controlling the mechanical arm to carry the punched sheet to the conveying start end of the transfer platform;

[0105] Step S4, controlling the transfer table to resume conveyance according to the image acquisition time of the punched film.

[0106] In implementation, steps S1 to S4 are performed consecutively, and between each step S3 and S4, the packaged folder is sent to the image segmentation module; the image segmentation module divides each punched sheet surface image into a number of surface area images according to the image cutting rule and generates corresponding secondary numbers; the surface area images with the same secondary number in the folder are fused to form a corresponding fused area image; the image analysis module determines the similarity of the fused area images in a single folder to determine whether the punched sheet corresponding to the folder is qualified.

[0107] The above description is only a preferred specific implementation manner of the present application, and is not intended to limit the protection scope of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A punching sheet stacking detection system based on visual monitoring, characterized in that: include: A mechanical transmission module, including a transfer platform and a mechanical arm, wherein the mechanical arm is used to carry the punched sheet to the conveying start end of the transfer platform, and the transfer platform is used to convey the punched sheet from the conveying start end to the detection position; an image acquisition module, including a fill light unit for providing a plurality of shooting light sources at different angles, an image acquisition unit for acquiring a surface image of the punched sheet, and a shooting control unit for controlling the fill light unit to sequentially turn on the shooting light sources at various angles and controlling the image acquisition unit to perform image acquisition and pack a plurality of punched sheet surface images of a single punched sheet into a folder and generate a primary number of each punched sheet surface image according to the image acquisition sequence; a transmission control module, which is connected to the mechanical transmission module and the image acquisition module respectively, and is used to control the transfer platform to stop transmission and control the image acquisition module to start image acquisition according to the pressure data of the detection position and the transmission time, and to control the transfer platform to resume transmission according to the image acquisition duration; an image segmentation module, which is connected to the image acquisition module, and is used to divide each punching sheet surface image into a plurality of surface area images according to the image cutting rule and generate corresponding secondary numbers, and fuse the surface area images with the same secondary number in the folder to form a corresponding fused area image; An image analysis module is connected to the image segmentation module, and compares each of the fusion area images in a single folder with the corresponding standard secondary images to determine the similarity of each fusion area image, and determines whether the punched film corresponding to the folder is qualified based on the similarity of each fusion area image; wherein, in a single image acquisition, the number of punched film surface images generated is equal to the number of shooting light sources turned on, the surface area image includes a center area image and an edge area image, and the standard secondary image is a surface area image of a punched film standard part at the same position.

2. A punching sheet stacking detection system based on visual monitoring according to claim 1, characterized in that: The detection position is located on a base of the transfer platform, and a pressure sensor for determining pressure data of the transfer platform on the detection position is arranged between the detection position and the transfer platform.

3. The punching sheet stacking detection system based on visual monitoring according to claim 1, characterized in that: The transmission control module determines whether the punch has reached the detection position based on the pressure data and transmission time at the detection position, wherein if the pressure data is within a reference pressure range and the transmission time is within a preset time range, it is determined that the punch has reached the detection position; if the pressure data is not within the reference pressure range and / or the transmission time is not within the preset time range, it is determined that the punch has not reached the detection position.

4. The punching sheet stacking detection system based on visual monitoring according to claim 3 is characterized in that: The conveying control module controls the transfer table to stop conveying and controls the image acquisition module to start image acquisition according to the determination result that the punching sheet has arrived at the detection position; wherein, controlling the image acquisition module to start image acquisition includes the shooting control unit controlling the fill light unit to turn on the shooting light sources at various angles in sequence and controlling the image acquisition unit to perform image acquisition when turning on the shooting light sources at various angles.

5. The punching sheet stacking detection system based on visual monitoring according to claim 1, characterized in that: The conveying control module determines the end of image acquisition of a single sheet according to the image acquisition time of the single sheet and controls the transfer table to resume conveying. If the image acquisition time of a single sheet is equal to the preset time, it is determined that the image acquisition of the sheet is ended and the transfer table is controlled to resume conveying.

6. The punching sheet stacking detection system based on visual monitoring according to claim 1, characterized in that: The image segmentation module pre-stores image cutting rules, which include determining, based on a reference area ratio, to divide each punching surface image into an edge image and a center image, and determining, based on a detection accuracy, a first reference number and a second reference number to respectively divide the edge image and the center image into a first reference number of edge area images and a second reference number of center area images; wherein the center image and the corresponding punching surface image are similar images, and the geometric center of the center image is the geometric center of the corresponding punching surface image.

7. The punching sheet stacking detection system based on visual monitoring according to claim 6, characterized in that: The reference area ratio is the area ratio of the center image to the image on the punching sheet surface.

8. The punching sheet stacking detection system based on visual monitoring according to claim 1, characterized in that: The image analysis module is preset with a machine learning model to determine the similarity of the images in each fusion area.

9. The punching sheet stacking detection system based on visual monitoring according to claim 1, characterized in that: The image analysis module determines whether the punched film corresponding to the folder is qualified based on the similarity of each fusion area image, including determining whether the corresponding punched film has missing corners based on the similarity of the fusion area images corresponding to each edge area image, and determining whether the corresponding punched film has defects based on the similarity of the fusion area images corresponding to each center area image.

10. A punching sheet overlap detection method based on visual monitoring, using the punching sheet overlap detection system based on visual monitoring according to any one of claims 1 to 9, characterized in that: include: Step S1, obtaining pressure data and transmission time data of the detection position in real time; Step S2, determining whether the punching sheet has reached the detection position based on the pressure data and transmission time of the detection position; Step S3, controlling the transfer platform to stop conveying and controlling the image acquisition module to start image acquisition based on the judgment result of reaching the detection position, and controlling the robotic arm to move the punching sheet to the conveying starting end of the transfer platform; Step S4, controlling the transfer platform to resume conveying based on the image acquisition time of the punching sheet.

Images