A method for real-time detection of road coating defects and related products
Patent Information
- Application Number
- CN202611123739.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-27
- Publication Date
- 2026-10-16
AI Technical Summary
随着产线自动化程度的日益提升，使用机械臂或运动平台在指定路径上涂覆胶体，难免由于运动抖动、出胶轨迹误差、胶枪存在气泡等因素干扰而出现缺陷，如断胶、胶宽不均匀、位置偏差等
[0016]从以上技术方案可以看出，相较于现有技术，本申请具有以下优点：
Smart Images
Figure CN122928797A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of industrial testing technology, and in particular to a method and related products for real-time detection of defects in coated adhesive lines. Background Technology
[0002] In automotive electronics manufacturing, applying thermal adhesive to product surfaces is a crucial step in extending product lifespan. With increasing automation in production lines, using robotic arms or motion platforms to apply adhesive along designated paths inevitably leads to defects such as adhesive breaks, uneven adhesive width, and positional deviations due to factors like motion vibration, adhesive trajectory errors, and air bubbles in the glue gun.
[0003] Existing adhesive path inspection methods typically rely on machine vision. However, most machine vision-based adhesive path inspection methods scan and inspect using preset paths or adhesive path templates. Scanning the entire adhesive path is time-consuming, and manual re-teaching is required when the coating process and path are changed. In addition, this method is susceptible to interference from ambient light or requires adjustments to the adhesive process parameters, necessitating adjustments to the algorithm parameters by skilled technicians, which leads to low defect detection efficiency.
[0004] Therefore, how to improve the efficiency of defect detection is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention
[0005] To address the aforementioned issues, this application provides a method and related products for real-time detection of defects in coated adhesive paths. The method directly determines the target adhesive to be detected based on the original adhesive path image, and then performs adhesive breakage detection and adhesive width analysis on the target adhesive to achieve real-time detection of defects in coated adhesive paths. Compared with machine vision-based adhesive path detection methods, the method does not require algorithm readjustment when affected by ambient light or when adjusting adhesive process parameters, thus improving defect detection efficiency.
[0006] In a first aspect, embodiments of this application provide a method for real-time detection of defects in coated adhesive lines, including: Obtain the original adhesive path image corresponding to the target adhesive path; The original adhesive path image is processed to determine the target adhesive in the original adhesive path image; The target colloid is subjected to target analysis, and the corresponding analysis results are obtained; the target analysis includes colloid breakage detection and colloid width analysis; Based on the analysis results, the detection results corresponding to the target adhesive coating path are determined and the detection results are output.
[0007] Optionally, processing the original adhesive path image and determining the target adhesive in the original adhesive path image includes: The original adhesive path image is corrected using a preset perspective transformation matrix to obtain the target corrected image; The target correction image is segmented using a preset adhesive path segmentation model to obtain the adhesive head segmentation result; The location of the adhesive path to be tested is determined based on the adhesive head segmentation results. The target colloid is determined from the target calibration image based on the location of the adhesive path to be tested.
[0008] Optionally, the step of performing target analysis on the target colloid and obtaining the corresponding analysis results includes: The target colloid was subjected to a breakage test, and a first analytical result was obtained; Extract the main framework of the target colloid and determine the colloid edge based on the main framework; Based on the edge of the colloid, determine the width value at each position of the target colloid; The width value is analyzed in conjunction with a first preset threshold, and a second analysis result is obtained.
[0009] Optionally, the method further includes: Acquire an actual marker image and a target marker image; the actual marker image is an image corresponding to a preset marker acquired when the coating head is in the coating pose; the target marker image is an image corresponding to a preset marker acquired when the coating head is perpendicular to the preset marker. Extract corresponding corner point pairs of preset marks from the actual marked image and the target marked image; A preset perspective transformation matrix is determined based on the corresponding corner point pairs.
[0010] Optionally, the method further includes: A training sample set is constructed based on data augmentation tools; the training sample set contains n sample glue path images labeled with glue path, glue head and background; different sample glue path images differ in at least one of the following: acquisition angle, lighting conditions and glue path shape; n is an integer greater than or equal to 2; The semantic segmentation model is trained based on the training sample set, and a preset glue path segmentation model is obtained.
[0011] Optionally, determining the location of the adhesive path to be tested based on the adhesive head segmentation result includes: The position of the rubber head is determined based on the rubber head segmentation results; Establish a search range corresponding to the adhesive path to be tested based on the position of the adhesive head; The search range is intersected with each segment of the glue path in the glue head segmentation result, and the position of the glue path segment with an intersection ratio greater than the second preset threshold is taken as the position of the glue path to be tested.
[0012] Optionally, the step of performing a gel breakage detection on the target colloid and obtaining a first analytical result includes: The target colloids are classified to obtain a first-class colloid set and a second-class colloid set; the target colloids in the first-class colloid set are all straight colloids, and the target colloids in the second-class colloid set are all curved colloids. The coating extension directions corresponding to the target colloids in the first type of adhesive path concentration and the target colloids in the second type of adhesive path concentration are obtained respectively; Based on the coating extension direction, the target colloids in the first type of adhesive path concentration and each target colloid in the second type of adhesive path concentration are projected, and the corresponding projected areas are obtained. The adhesive breakage status of each target colloid is determined by combining the projected area and the third preset threshold, and the adhesive breakage status is used as the first analysis result.
[0013] Secondly, embodiments of this application provide a device for real-time detection of defects in coated adhesive lines, comprising: The data acquisition module is used to acquire the original adhesive path image corresponding to the target adhesive path. The data processing module is used to process the original adhesive path image and determine the target adhesive in the original adhesive path image; The data analysis module is used to perform target analysis on the target colloid and obtain corresponding analysis results; the target analysis includes colloid breakage detection and colloid width analysis. The output module is used to determine the detection result corresponding to the target coating adhesive path based on the analysis result, and to output the detection result.
[0014] Thirdly, embodiments of this application provide a device for real-time detection of defects in coated adhesive lines, comprising: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the method for real-time detection of defects in coated adhesive lines as described above.
[0015] Fourthly, embodiments of this application provide a readable storage medium, characterized in that the readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for real-time detection of defects in coated adhesive paths as described above.
[0016] As can be seen from the above technical solutions, compared with the prior art, this application has the following advantages: This application provides a method for real-time detection of defects in coated adhesive paths, comprising: first, acquiring an original adhesive path image corresponding to the target coated adhesive path; then, processing the original adhesive path image and identifying the target adhesive in the original adhesive path image; next, performing target analysis on the target adhesive and obtaining corresponding analysis results, wherein the target analysis includes adhesive breakage detection and adhesive width analysis; finally, determining the detection result corresponding to the target coated adhesive path based on the analysis results and outputting the detection result. Thus, by directly identifying the target adhesive to be detected based on the original adhesive path image, and then performing adhesive breakage detection and adhesive width analysis on the target adhesive, real-time detection of defects in coated adhesive paths can be achieved. Compared with machine vision-based adhesive path detection methods, this method eliminates the need to readjust the algorithm when affected by ambient light interference or when adjusting adhesive process parameters, thereby improving defect detection efficiency. Attached Figure Description
[0017] Figure 1 A flowchart illustrating a method for real-time detection of defects in coated adhesive lines, provided in an embodiment of this application; Figure 2 A schematic diagram of a projection method for detecting adhesive breakage provided in an embodiment of this application; Figure 3 This application provides a schematic diagram of the extraction of the main skeleton of the adhesive path. Figure 4 A schematic diagram of a caliper rectangular frame for detecting adhesive path width provided in an embodiment of this application; Figure 5 A schematic diagram of a captured marker image provided in an embodiment of this application; Figure 6 A schematic diagram of a semantic segmentation model label provided in an embodiment of this application; Figure 7 This is a schematic diagram of a device for real-time detection of defects in coated adhesive lines, provided in an embodiment of this application. Detailed Implementation
[0018] As mentioned earlier, existing adhesive path detection methods suffer from low defect detection efficiency. Specifically, these methods typically rely on machine vision, which mostly scans pre-defined paths or adhesive path templates. Scanning the entire adhesive path is time-consuming, and manual re-teaching is required when the coating process or path is changed. Furthermore, these methods are susceptible to interference from ambient light or require adjustments to adhesive process parameters, necessitating adjustments by skilled technicians, further contributing to the low defect detection efficiency.
[0019] To address the aforementioned problems, embodiments of this application provide a method for real-time detection of defects in coated adhesive paths, comprising: first, acquiring an original adhesive path image corresponding to the target coated adhesive path; then, processing the original adhesive path image and identifying the target adhesive within the image; next, performing target analysis on the target adhesive and obtaining corresponding analysis results, wherein the target analysis includes adhesive breakage detection and adhesive width analysis; finally, determining the detection result corresponding to the target coated adhesive path based on the analysis results and outputting the detection result.
[0020] In this way, the target adhesive to be detected is determined directly based on the original adhesive path image, and then the adhesive breakage detection and adhesive width analysis are performed on the target adhesive to achieve real-time detection of coating adhesive path defects. Compared with the adhesive path detection method based on machine vision, there is no need to readjust the algorithm when affected by ambient light or when adjusting the adhesive process parameters, thus improving the defect detection efficiency.
[0021] It should be noted that the method and related products for real-time detection of defects in coated adhesive lines provided in this application can be applied to the field of industrial inspection technology. The above are merely examples and do not limit the application areas of the method and related products for real-time detection of defects in coated adhesive lines provided in this application.
[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0023] Figure 1 This is a flowchart illustrating a method for real-time detection of defects in coated adhesive lines, provided as an embodiment of this application. (Combined with...) Figure 1 As shown in the figure, this application provides a method for real-time detection of defects in coated adhesive lines, which may include: S101: Obtain the original adhesive path image corresponding to the target adhesive path.
[0024] In practical applications, the real-time detection method for adhesive coating defects is jointly implemented by a data acquisition module, a data management module, and a data detection module. The data detection module further includes a data processing module, a data analysis module, and an output module. To improve product lifespan, a thermal adhesive is applied to the product surface; the adhesive coating path is the target adhesive coating path. During the coating process, the data acquisition module triggers a camera to acquire images of the coated adhesive path at the current location and stores them in the data management module. Essentially, during coating, the camera moves with the coating equipment, acquiring images of the adhesive path at various locations within the target adhesive coating path; these images are referred to as the original adhesive path images corresponding to the target adhesive coating path. To maximize detection efficiency, the data acquisition module, data management module, and detection module run in different threads, managing data storage and distribution through queues. Furthermore, by utilizing the overlap between adjacent frames and appropriately matching the data acquisition interval with the coating rate, a "real-time detection" effect of simultaneous coating and detection can be achieved.
[0025] S102: Process the original adhesive path image and determine the target adhesive in the original adhesive path image.
[0026] In practical applications, when the data detection module is idle, the data management module sends the original adhesive path image to the data detection module for inspection. Since the original adhesive path image may contain distortions or interference with the adhesive path, the data detection module first needs to preprocess the original adhesive path image upon receiving it. Then, the target adhesive to be detected is extracted from the processed adhesive path image for subsequent adhesive path defect detection.
[0027] Furthermore, since the methods for determining the target colloid are not entirely the same, the embodiments of this application can describe one possible method of determination.
[0028] In one scenario, S102: Processing the original adhesive path image and determining the target colloid in the original adhesive path image may specifically include: The original adhesive path image is corrected using a preset perspective transformation matrix to obtain the target corrected image; The target correction image is segmented using a preset adhesive path segmentation model to obtain the adhesive head segmentation result; The location of the adhesive path to be tested is determined based on the adhesive head segmentation results. The target colloid is determined from the target calibration image based on the location of the adhesive path to be tested.
[0029] In practical applications, the original glue path image first needs to undergo perspective correction, glue path segmentation, and glue path selection to obtain the target glue in the original glue path image. Specifically, during perspective correction, a calibrated perspective transformation matrix (preset perspective transformation matrix) can be introduced to correct the original glue path image, obtaining an image from a directly overhead view, which is denoted as the target corrected image (warp_image). Then, a trained glue path segmentation model (preset glue path segmentation model) is used to segment the target corrected image, separating various glue paths, glue heads, and background, which is used as the glue head segmentation result (glue_mask). Furthermore, since interfering glue paths inevitably exist in each original glue path image, it is also necessary to select the location of the glue path to be detected from the glue head segmentation result corresponding to the original glue path image, and record it as the glue path location to be tested (selected_glue_mask). Finally, the glue part is extracted from the target corrected image using the glue path location to be tested, recorded as the target glue (selected_glue_image), and used as the subsequent detection target.
[0030] Furthermore, since the methods for determining the location of the adhesive path to be tested are not entirely the same, the embodiments of this application can describe one possible determination method.
[0031] In one scenario, determining the location of the adhesive path to be tested based on the adhesive head segmentation result includes: The position of the rubber head is determined based on the rubber head segmentation results; Establish a search range corresponding to the adhesive path to be tested based on the position of the adhesive head; The search range is intersected with each segment of the glue path in the glue head segmentation result, and the position of the glue path segment with an intersection ratio greater than the second preset threshold is taken as the position of the glue path to be tested.
[0032] In practical applications, due to the large field of view of the camera, interference from historical adhesive paths and environmental factors is inevitable when acquiring original adhesive path images. Therefore, it is necessary to filter out the target adhesives for detection from the target calibration images. Specifically, firstly, the adhesive head needs to be extracted from the adhesive head segmentation results and its position determined. Then, a search range (search_mask) for the adhesive path to be detected is established based on the adhesive head position. Understandably, the shape of the search range can be adjusted according to the pattern of the coated adhesive path. Further, the intersection of the search range and each segment of the adhesive path in the adhesive head segmentation results is calculated. When the intersection ratio (intersection area / adhesive path area) is greater than a second preset threshold (e.g., 0.6, 0.7, 0.8, etc.), the selected adhesive path segment is considered to be within the search range. The position of the adhesive path segment that meets the condition is recorded as the position of the adhesive path to be tested. Then, the target adhesive is determined based on the position of the adhesive path to be tested. The set of all adhesive path segments within the search range is recorded as the target set (search_mask_list). In addition, if the intersection ratio of all glue segments is less than or equal to the second preset threshold, it means that there is no effective glue path within the search range. In this case, there is no need to perform subsequent glue breakage analysis, and the detection result of "no effective glue path" is returned directly.
[0033] S103: Perform target analysis on the target colloid and obtain the corresponding analysis results; the target analysis includes colloid breakage detection and colloid width analysis.
[0034] In practical applications, after identifying the target colloid, the data detection module also needs to perform colloid breakage detection and colloid width analysis step by step to obtain the corresponding analysis results.
[0035] Furthermore, since the methods for analyzing target colloids are not entirely the same, the embodiments of this application can be used to describe one possible analytical method.
[0036] In one scenario, S103: Target analysis is performed on the target colloid, and corresponding analysis results are obtained, which may specifically include: The target colloid was subjected to a breakage test, and a first analytical result was obtained; Extract the main framework of the target colloid and determine the colloid edge based on the main framework; Based on the edge of the colloid, determine the width value at each position of the target colloid; The width value is analyzed in conjunction with a first preset threshold, and a second analysis result is obtained.
[0037] In practical applications, the analysis of the target colloid involves two aspects. First, it involves detecting colloid breakage to obtain the first analytical result. This first result indicates whether colloid breakage exists and its location. Second, it involves analyzing the colloid width to obtain the second analytical result. For width analysis, the main framework of the target colloid is extracted first. Then, a series of calipers are placed at equal intervals along the framework, and the colloid edge is determined based on these calipers. The width value at each location is then determined based on the colloid edge. Finally, the width value at each location is compared with a first preset threshold to determine if the colloid width at each location meets the requirements. The second analytical result then displays the locations of colloids that meet the width requirements and those that do not.
[0038] Furthermore, since the analytical methods for detecting glue breakage are not entirely the same, the embodiments of this application can be used to describe one possible analytical method.
[0039] In one scenario, the step of performing a gel breakage detection on the target colloid and obtaining a first analytical result includes: The target colloids are classified to obtain a first-class colloid set and a second-class colloid set; the target colloids in the first-class colloid set are all straight colloids, and the target colloids in the second-class colloid set are all curved colloids. The coating extension directions corresponding to the target colloids in the first type of adhesive path concentration and the target colloids in the second type of adhesive path concentration are obtained respectively; Based on the coating extension direction, the target colloids in the first type of adhesive path concentration and each target colloid in the second type of adhesive path concentration are projected, and the corresponding projected areas are obtained. The adhesive breakage status of each target colloid is determined by combining the projected area and the third preset threshold, and the adhesive breakage status is used as the first analysis result.
[0040] In practical applications, the target colloids in the target set can be divided into two categories based on their segmentation labels: straight adhesive paths and curved adhesive paths. The set of all target colloids forming straight adhesive paths can be called the first-class adhesive path set (merge_straight_mask), while the set of all target colloids forming curved adhesive paths can be called the second-class adhesive path set (merge_curved_mask). Furthermore, a projection adhesive breakage detection step is performed on each target colloid within both the first-class and second-class adhesive path sets to determine whether any adhesive breakage exists. Figure 2 This is a schematic diagram of a projection method for detecting adhesive breakage provided in an embodiment of this application. (In conjunction with...) Figure 2As shown, the detection steps can be as follows: First, obtain the coating extension direction (main direction) corresponding to the adhesive path map (target adhesive). Then, project the target adhesive in a direction perpendicular to the main direction. When the projected area (the area of the adhesive in each column of pixels) in the projection direction reaches a threshold, the projection map is recorded as 1; otherwise, it is recorded as 0. The length of discontinuous 0s in the projection map is calculated. If the length is greater than a third preset threshold (e.g., 1 pixel, 2 pixels, etc.), it is considered that there is a break in the adhesive, and the location of the break is recorded.
[0041] Furthermore, regarding how to achieve the width value analysis of the target colloid, the embodiments of this application can describe one possible width value analysis method.
[0042] In one scenario, to analyze the width of the adhesive path, it is necessary to first extract the main skeleton of each target colloid in the target set, and then record the single-pixel coordinates (skeleton_pts) of the main skeleton through distance transformation, thinning, morphological erosion, etc. Next, it is determined whether there are branches on the main skeleton; if so, branch removal is performed, and the node coordinates of non-main branches are removed from the single-pixel coordinates. Figure 3 This is a schematic diagram of the extraction of the main skeleton of the adhesive path provided in an embodiment of this application. Figure 4 This is a schematic diagram of a caliper rectangular frame for measuring adhesive path width, provided in an embodiment of this application. (Combined with...) Figure 3 and Figure 4 As shown, firstly, the connectivity value of each node within the main skeleton (i.e., the number of branches passing through the node) is calculated, and nodes with a connectivity value greater than 2 are identified and recorded as branch nodes. Then, the main skeleton is divided into multiple segments based on the branch nodes, and the length of each segment is calculated. Using a depth-first search algorithm, combined with the connectivity of each branch, the longest connected skeleton segment is selected and merged to form the main branch. Non-main branches are then removed, resulting in the removal of small branches. Next, the normal direction of each node position is calculated within the single-pixel coordinates after removing the node coordinates on the non-main branches. Based on the detection configuration sampling points, a series of caliper rectangles are generated at equal intervals along the normal direction. The caliper rectangles are then used to extract the original adhesive path image, and the edge points on both sides of the internal colloid (target colloid) are detected. From this, the width value at each position of the target colloid can be calculated. Finally, based on the width values at each position and a first preset threshold, it is determined whether there are defects such as unqualified width or uneven width, and the corresponding second analysis result is obtained.
[0043] In addition, this application also provides a method for determining a preset perspective transformation matrix, which may specifically include: Acquire an actual marker image and a target marker image; the actual marker image is an image corresponding to a preset marker acquired when the coating head is in the coating pose; the target marker image is an image corresponding to a preset marker acquired when the coating head is perpendicular to the preset marker. Extract corresponding corner point pairs of preset markers from the actual marked image and the target marked image; A preset perspective transformation matrix is determined based on the corresponding corner point pairs.
[0044] In practical applications, the coating head may not be perfectly perpendicular to the workpiece surface, resulting in a fixed angle between the camera and the target coating path, which will distort the obtained original coating path image. Therefore, to ensure measurement accuracy, it is necessary to set a reasonable perspective transformation matrix and perform perspective correction on the acquired original coating path image based on this matrix. Figure 5 This is a schematic diagram illustrating the acquisition of a marked image according to an embodiment of this application. (In conjunction with...) Figure 5 As shown, a checkerboard pattern can be placed on the workpiece plane as a marker. A laser level is used to assist in aligning the camera plane, fixed above the adhesive applicator, with the camera plane parallel to the workpiece plane. An image is then acquired and recorded as the target marker image (calib_image). During leveling, marker images are continuously acquired. The parallelism of the camera plane to the workpiece plane is determined by whether the grid size in the acquired marker image is square. If not, the camera plane needs fine-tuning until the grid size is square; at this point, the camera plane is considered parallel to the workpiece plane. Then, the adhesive applicator is adjusted to the adhesive application position, and the checkerboard image is acquired again to obtain the actual marker image (real_image). Finally, corresponding corner point pairs of the checkerboard pattern are extracted from the two images, and the perspective transformation matrix from the actual marker image to the target marker image is calculated. This calculated perspective transformation matrix is used as the calibrated perspective transformation matrix (preset perspective transformation matrix) to correct each original adhesive path image.
[0045] In addition, embodiments of this application also provide a method for obtaining a preset adhesive path segmentation model, which may specifically include: A training sample set is constructed based on data augmentation tools; the training sample set contains n sample glue path images labeled with glue path, glue head and background; different sample glue path images differ in at least one of the following: acquisition angle, lighting conditions and glue path shape; n is an integer greater than or equal to 2; The semantic segmentation model is trained based on the training sample set, and a preset glue path segmentation model is obtained.
[0046] In practical applications, due to the potential for multiple interferences during adhesive coating processes, such as variations in adhesive shape, color, and ambient light, this application introduces a segmentation model to replace the traditional threshold segmentation method, thereby improving the algorithm's robustness. During model training, n adhesive path images with different angles, lighting conditions, or shapes are collected as sample adhesive path images. Here, n is an integer greater than or equal to 2. Figure 6 This is a schematic diagram of a semantic segmentation model label provided in an embodiment of this application. Combined with... Figure 6 As shown, each sample glue path image is labeled with the glue path, glue head, and background. To improve the robustness of subsequent detection, straight and curved regions of the glue path are further labeled as straight glue path and curved glue path, respectively. Finally, by combining all the sample glue path images, a training sample set is generated using data augmentation tools, and the semantic segmentation model is trained to obtain a trained glue path segmentation model (the preset glue path segmentation model). Alternatively, models such as PIDNet, DDRNet, and UNet can be selected.
[0047] S104: Based on the analysis results, determine the detection results corresponding to the target adhesive coating path, and output the detection results.
[0048] In practical applications, the analysis results obtained by performing target analysis on each original adhesive path image are sent to the data management module for temporary storage. After all adhesive paths have been coated and all corresponding original adhesive path images have been analyzed, the final detection result corresponding to the entire coated adhesive path (target coated adhesive path) is determined by combining the analysis results corresponding to each original adhesive path image, and the final detection result is output.
[0049] In summary, this application first acquires the original adhesive path image corresponding to the target coated adhesive path. Then, the original adhesive path image is processed to identify the target adhesive within it. Next, target analysis is performed on the target adhesive, yielding corresponding analysis results, including adhesive breakage detection and adhesive width analysis. Finally, based on the analysis results, the detection result corresponding to the target coated adhesive path is determined and output. Thus, by directly identifying the target adhesive based on the original adhesive path image, and then performing adhesive breakage detection and adhesive width analysis on the target adhesive, real-time detection of defects in coated adhesive paths can be achieved. Compared to machine vision-based adhesive path detection methods, this approach eliminates the need to readjust the algorithm when affected by ambient light interference or adjustments to adhesive process parameters, thereby improving defect detection efficiency.
[0050] Figure 7 This is a schematic diagram of a device for real-time detection of defects in coated adhesive lines, provided as an embodiment of this application. (Combined with...) Figure 7 As shown, the device 700 for real-time detection of defects in the coated adhesive path may include: Data acquisition module 701 is used to acquire the original adhesive path image corresponding to the target adhesive path; The data processing module 702 is used to process the original adhesive path image and determine the target adhesive in the original adhesive path image; The data analysis module 703 is used to perform target analysis on the target colloid and obtain corresponding analysis results; the target analysis includes colloid breakage detection and colloid width analysis. The output module 704 is used to determine the detection result corresponding to the target coating adhesive path based on the analysis result, and to output the detection result.
[0051] As one implementation method, the data processing module 702 is specifically used for determining the target colloid.
[0052] The original adhesive path image is corrected using a preset perspective transformation matrix to obtain the target corrected image; The target correction image is segmented using a preset adhesive path segmentation model to obtain the adhesive head segmentation result; The location of the adhesive path to be tested is determined based on the adhesive head segmentation results. The target colloid is determined from the target calibration image based on the location of the adhesive path to be tested.
[0053] Furthermore, determining the location of the adhesive path to be tested based on the adhesive head segmentation result includes: The position of the rubber head is determined based on the rubber head segmentation results; Establish a search range corresponding to the adhesive path to be tested based on the position of the adhesive head; The search range is intersected with each segment of the glue path in the glue head segmentation result, and the position of the glue path segment with an intersection ratio greater than the second preset threshold is taken as the position of the glue path to be tested.
[0054] As one implementation method, regarding how to perform target analysis on the target colloid, the aforementioned data analysis module 703 is specifically used for: The target colloid was subjected to a breakage test, and a first analytical result was obtained; Extract the main framework of the target colloid and determine the colloid edge based on the main framework; Based on the edge of the colloid, determine the width value at each position of the target colloid; The width value is analyzed in conjunction with a first preset threshold, and a second analysis result is obtained.
[0055] Furthermore, the step of performing a breakage detection on the target colloid and obtaining a first analytical result includes: The target colloids are classified to obtain a first-class colloid set and a second-class colloid set; the target colloids in the first-class colloid set are all straight colloids, and the target colloids in the second-class colloid set are all curved colloids. The coating extension directions corresponding to the target colloids in the first type of adhesive path concentration and the target colloids in the second type of adhesive path concentration are obtained respectively; Based on the coating extension direction, the target colloids in the first type of adhesive path concentration and each target colloid in the second type of adhesive path concentration are projected, and the corresponding projected areas are obtained. The adhesive breakage status of each target colloid is determined by combining the projected area and the third preset threshold, and the adhesive breakage status is used as the first analysis result.
[0056] As one implementation, the above-mentioned device 700 for real-time detection of coating adhesive defects further includes a determination module for determining the preset perspective transformation matrix; The determination module is used to acquire the actual mark image and the target mark image; the actual mark image is the image corresponding to the preset mark acquired when the coating head is in the coating pose; the target mark image is the image corresponding to the preset mark acquired when the coating head is perpendicular to the preset mark. Extract corresponding corner point pairs of preset markers from the actual marked image and the target marked image; A preset perspective transformation matrix is determined based on the corresponding corner point pairs.
[0057] As one implementation method, the above-mentioned device 700 for real-time detection of coating adhesive defects further includes a training module to address how to obtain a preset adhesive path segmentation model; The training module is used to construct a training sample set based on data augmentation tools; the training sample set contains n sample glue path images labeled with glue path, glue head and background; different sample glue path images differ in at least one of the following: acquisition angle, lighting conditions and glue path shape; n is an integer greater than or equal to 2; The semantic segmentation model is trained based on the training sample set, and a preset glue path segmentation model is obtained.
[0058] In summary, this application first acquires the original adhesive path image corresponding to the target coated adhesive path. Then, the original adhesive path image is processed to identify the target adhesive within it. Next, target analysis is performed on the target adhesive, yielding corresponding analysis results, including adhesive breakage detection and adhesive width analysis. Finally, based on the analysis results, the detection result corresponding to the target coated adhesive path is determined and output. Thus, by directly identifying the target adhesive based on the original adhesive path image, and then performing adhesive breakage detection and adhesive width analysis on the target adhesive, real-time detection of defects in coated adhesive paths can be achieved. Compared to machine vision-based adhesive path detection methods, this approach eliminates the need to readjust the algorithm when affected by ambient light interference or adjustments to adhesive process parameters, thereby improving defect detection efficiency.
[0059] In addition, this application also provides a device for real-time detection of defects in coated adhesive lines, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the method for real-time detection of defects in coated adhesive lines as described above.
[0060] In addition, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for real-time detection of defects in coated adhesive lines as described above.
[0061] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for real-time detection of defects in coated adhesive lines, characterized in that, The method includes: Obtain the original adhesive path image corresponding to the target adhesive path; The original adhesive path image is processed to determine the target adhesive in the original adhesive path image; The target colloid is subjected to target analysis, and the corresponding analysis results are obtained; the target analysis includes colloid breakage detection and colloid width analysis; Based on the analysis results, the detection results corresponding to the target adhesive coating path are determined and the detection results are output.
2. The method according to claim 1, characterized in that, The step of processing the original adhesive path image and determining the target adhesive in the original adhesive path image includes: The original adhesive path image is corrected using a preset perspective transformation matrix to obtain the target corrected image; The target correction image is segmented using a preset adhesive path segmentation model to obtain the adhesive head segmentation result; The location of the adhesive path to be tested is determined based on the adhesive head segmentation results. The target colloid is determined from the target calibration image based on the location of the adhesive path to be tested.
3. The method according to claim 1, characterized in that, The step of performing target analysis on the target colloid and obtaining the corresponding analysis results includes: The target colloid was subjected to a breakage test, and a first analytical result was obtained; Extract the main framework of the target colloid and determine the colloid edge based on the main framework; Based on the edge of the colloid, determine the width value at each position of the target colloid; The width value is analyzed in conjunction with a first preset threshold, and a second analysis result is obtained.
4. The method according to claim 1, characterized in that, The method further includes: Acquire an actual marker image and a target marker image; the actual marker image is an image corresponding to a preset marker acquired when the coating head is in the coating pose; the target marker image is an image corresponding to a preset marker acquired when the coating head is perpendicular to the preset marker. Extract corresponding corner point pairs of preset marks from the actual marked image and the target marked image; A preset perspective transformation matrix is determined based on the corresponding corner point pairs.
5. The method according to claim 1, characterized in that, The method further includes: A training sample set is constructed based on data augmentation tools; the training sample set contains n sample glue path images labeled with glue path, glue head and background; different sample glue path images differ in at least one of the following: acquisition angle, lighting conditions and glue path shape; n is an integer greater than or equal to 2; The semantic segmentation model is trained based on the training sample set, and a preset glue path segmentation model is obtained.
6. The method according to claim 2, characterized in that, The step of determining the location of the adhesive path to be tested based on the adhesive head segmentation result includes: The position of the rubber head is determined based on the rubber head segmentation results; Establish a search range corresponding to the adhesive path to be tested based on the position of the adhesive head; The search range is intersected with each segment of the glue path in the glue head segmentation result, and the position of the glue path segment with an intersection ratio greater than the second preset threshold is taken as the position of the glue path to be tested.
7. The method according to claim 3, characterized in that, The step of performing a breakage detection on the target colloid and obtaining a first analytical result includes: The target colloids are classified to obtain a first-class colloid set and a second-class colloid set; the target colloids in the first-class colloid set are all straight colloids, and the target colloids in the second-class colloid set are all curved colloids. The coating extension directions corresponding to the target colloids in the first type of adhesive path concentration and the target colloids in the second type of adhesive path concentration are obtained respectively; Based on the coating extension direction, the target colloids in the first type of adhesive path concentration and each target colloid in the second type of adhesive path concentration are projected, and the corresponding projected areas are obtained. The adhesive breakage status of each target colloid is determined by combining the projected area and the third preset threshold, and the adhesive breakage status is used as the first analysis result.
8. A device for real-time detection of defects in coated adhesive lines, characterized in that, include: The data acquisition module is used to acquire the original adhesive path image corresponding to the target adhesive path. The data processing module is used to process the original adhesive path image and determine the target adhesive in the original adhesive path image; The data analysis module is used to perform target analysis on the target colloid and obtain the corresponding analysis results; The target analysis includes adhesive breakage detection and colloid width analysis; The output module is used to determine the detection result corresponding to the target coating adhesive path based on the analysis result, and to output the detection result.
9. A device for real-time detection of defects in coated adhesive lines, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the method for real-time detection of defects in coated adhesive paths as described in any one of claims 1 to 7.
10. A readable storage medium, characterized in that, The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method for real-time detection of defects in coated adhesive lines as described in any one of claims 1 to 7.