Photovoltaic product detection method and device, storage medium and computer equipment

Abstract

The application discloses a photovoltaic product detection method and device, a storage medium and computer equipment, relates to the photovoltaic detection technical field, and mainly aims at improving the detection efficiency and detection accuracy of photovoltaic products. The method comprises the following steps: acquiring a plurality of RGBD images of a to-be-detected photovoltaic product during movement along a production line in real time; performing image decomposition on each RGBD image respectively, and performing image contour extraction on the decomposed RGBD image; determining a line drawing of the to-be-detected photovoltaic product based on the extracted image contour; matching the line drawing with a standard line drawing in a preset photovoltaic product library, and determining a detection result of the to-be-detected photovoltaic product based on the matching result, wherein the preset photovoltaic product library stores standard line drawings corresponding to a plurality of qualified photovoltaic products. The application is suitable for the scene of quality detection of photovoltaic products.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic testing technology, and in particular to a testing method, apparatus, storage medium, and computer equipment for photovoltaic products. Background Technology

[0002] In industrial photovoltaic product production lines, quality inspections are required at multiple stages of production, including but not limited to photovoltaic panel inspection and glue filling inspection.

[0003] Currently, photovoltaic products are typically inspected manually. However, this manual inspection method is time-consuming and labor-intensive, and due to the negligence of staff and the varying levels of technical expertise, it can lead to missed or incorrect inspections. Summary of the Invention

[0004] This invention provides a method, apparatus, storage medium, and computer equipment for testing photovoltaic products, which mainly improves the testing efficiency and accuracy of photovoltaic products.

[0005] According to a first aspect of the present invention, a method for testing photovoltaic products is provided, comprising: Multiple RGBD images of the photovoltaic product under test are acquired in real time as it moves along the production line. Each RGBD image is decomposed, and the decomposed RGBD images are used to extract image contours. Based on the extracted image contours, the line drawing of the photovoltaic product to be detected is determined. The line drawing is matched with the standard line drawing in the preset photovoltaic product library. Based on the matching result, the test result of the photovoltaic product to be tested is determined. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

[0006] Optionally, the line diagram is a line diagram corresponding to RGBD images captured by multiple camera devices arranged alternately on both sides of the production line, including a left-side line diagram corresponding to multiple left-side RGBD images captured by a camera device arranged on the left side of the production line, and a right-side line diagram corresponding to multiple right-side RGBD images captured by a camera device arranged on the right side of the production line. Before matching the line drawing with a standard line drawing in a preset photovoltaic product library, the method further includes: Take either the left RGBD image or the right RGBD image as the target RGBD image, determine multiple current frame feature points in the current frame RGBD image of the target RGBD image, and determine multiple previous frame feature points in the previous frame RGBD image on the same side as the current frame RGBD image. Feature point matching is performed on the feature points of the current frame and the feature points of the previous frame to obtain successfully matched feature point pairs. Based on the position information of the feature point pairs, the pose transformation parameters of the photovoltaic product to be detected along the production line are determined. The three-dimensional point cloud of the previous frame of the line graph corresponding to the previous frame RGBD image and the three-dimensional point cloud of the current frame of the line graph corresponding to the current frame RGBD image are determined respectively. Based on the pose transformation parameters, the previous frame 3D point cloud is transformed into the current coordinate system where the current frame 3D point cloud is located, and the transformed previous frame 3D point cloud and the current frame 3D point cloud are fused to obtain a fused 3D point cloud. Determine the degree of matching between the fused 3D point cloud and the current frame 3D point cloud, and based on the degree of matching, select a target line image for subsequent matching from the left line image corresponding to the left RGBD image and the right line image corresponding to the right RGBD image. The step of matching the line drawing with a standard line drawing in a preset photovoltaic product library includes: The target line drawing is matched with the standard line drawing in the preset photovoltaic product library.

[0007] Optionally, determining the degree of matching between the fused 3D point cloud and the current frame 3D point cloud includes: Calculate the distance between each fusion point in the fused 3D point cloud and each current point in the current frame 3D point cloud, determine the minimum distance corresponding to each fusion point in each distance, and determine the maximum distance in the minimum distance corresponding to each fusion point as the maximum fusion distance; Calculate the distance between each current point in the current frame 3D point cloud and each fused point in the fused 3D point cloud, and determine the minimum distance corresponding to each current point in each distance, and determine the maximum distance in each minimum distance as the maximum current distance; Based on the maximum fusion distance and the maximum current distance, determine the global shape matching degree between the fused 3D point cloud and the current frame 3D point cloud; Based on the curvature values ​​of each point in the fused 3D point cloud and the curvature values ​​of each point in the current frame 3D point cloud, multiple curvature intervals are determined; The fusion probability density of the fused 3D point cloud in each curvature interval and the current probability density of the current frame 3D point cloud in each curvature interval are determined respectively. The fusion curvature evaluation value of the fused 3D point cloud is determined based on the fusion probability density, and the current curvature evaluation value of the current frame 3D point cloud is determined based on the current probability density. Based on the fused curvature evaluation value and the current curvature evaluation value, the local edge matching degree between the fused 3D point cloud and the current frame 3D point cloud is determined; Based on the global shape matching degree and the local edge matching degree, the degree of matching between the fused 3D point cloud and the current frame 3D point cloud is determined.

[0008] Optionally, the step of selecting a target line image for subsequent matching based on the matching degree, between the left line image corresponding to the left RGBD image and the right line image corresponding to the right RGBD image, includes: The motion feature data of the photovoltaic product to be detected between the previous frame RGBD image and the current frame RGBD image is obtained, wherein the motion feature data includes motion speed and vibration frequency; Based on the motion feature data, a preset matching degree threshold is dynamically determined; If the matching degree is greater than the preset matching degree threshold, then select any line drawing from the line drawing corresponding to the target RGBD image as the target line drawing; otherwise, select any line drawing from the line drawing corresponding to the other RGBD image corresponding to the target RGBD image as the target line drawing.

[0009] Optionally, matching the line drawing with a standard line drawing in a preset photovoltaic product library includes: A predetermined number of standard points are evenly distributed on the outer contour of the standard line drawing, and a detection point corresponding to each standard point is determined on the line drawing; Determine the point distance between each of the aforementioned standard points and the corresponding point to be detected; Determine the angle between two adjacent standard points in each of the standard points, and determine the angle between two adjacent test points corresponding to the two adjacent standard points; Based on the positional relationship between each standard point and its corresponding adjacent standard points, the standard extension direction angle corresponding to each standard point is determined; based on the positional relationship between each point to be detected and its corresponding adjacent points to be detected, the detection extension direction angle corresponding to each point to be detected is determined. Based on the point distance, the line drawing is matched with the standard line drawing in the preset photovoltaic product library to obtain a distance matching result. Based on the point angle and the angle of the point to be detected, the line drawing is matched with the standard line drawing in the preset photovoltaic product library to obtain an angle matching result. Based on the standard extension direction angle and the extension direction angle to be detected, the line drawing is matched with the standard line drawing in the preset photovoltaic product library to obtain an extension angle matching result. Based on the distance matching result, the included angle matching result, and the extension angle matching result, the matching result between the line drawing and the standard line drawing in the preset photovoltaic product library is determined.

[0010] Optionally, the step of extracting image contours from the decomposed RGBD image and determining the line drawing of the photovoltaic product to be detected based on the extracted image contours includes: Determine the depth map of the decomposed RGBD image; Extract the image edges of the decomposed RGBD image and determine whether there are elevation changes in the depth map. If so, the image edges are determined as the image contours of the decomposed RGBD image; otherwise, the image edges are determined as pseudo contours and filtered out. The image outline is decomposed into a variety of vector graphic primitives, wherein the variety of vector graphic primitives includes at least one of line segments, polylines, circles, arcs, ellipses, and text; Each of the vector graphic primitives is combined according to its original position in the image outline, and based on the combination result, a line drawing of the photovoltaic product to be tested is obtained.

[0011] Optionally, image decomposition is performed on each of the RGBD images, including: For each RGBD image, the RGB channels are separated to obtain multiple RGB channel images; Based on the pixel value distribution information of each RGB channel image, a color spot region is determined in the corresponding RGB channel image, and the color spot region is removed in the corresponding RGB channel image to obtain each despotted RGB channel image. Each despotted RGB channel image constitutes the RGBD image.

[0012] According to a second aspect of the present invention, a testing apparatus for photovoltaic products is provided, comprising: The acquisition unit is used to acquire multiple RGBD images of the photovoltaic product under test as it moves along the production line in real time. The determining unit is used to decompose each of the RGBD images, extract the image contours of the decomposed RGBD images, and determine the line drawing of the photovoltaic product to be detected based on the extracted image contours. The matching unit is used to match the line drawing with the standard line drawing in the preset photovoltaic product library, and determine the test result of the photovoltaic product to be tested based on the matching result. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

[0013] According to a third aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for detecting photovoltaic products.

[0014] According to a fourth aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for detecting photovoltaic products.

[0015] According to the present invention, a method, apparatus, storage medium, and computer equipment for testing photovoltaic products are provided. Compared with the current method of manually inspecting photovoltaic products, the present invention acquires multiple RGBD images of the photovoltaic product to be tested in real time as it moves along the production line. Then, each RGBD image is decomposed, and image contours are extracted from the decomposed RGBD images. Based on the extracted image contours, a line drawing of the photovoltaic product to be tested is determined. Finally, the line drawing is matched with standard line drawings in a preset photovoltaic product library. Based on the matching result, the test result of the photovoltaic product to be tested is determined. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products. Therefore, by extracting the line drawing of the photovoltaic product from the RGBD images and matching it with the standard line drawings in the preset photovoltaic product library to determine the test result, the time wasted and errors associated with manual inspection can be avoided. Thus, the present invention, through visual inspection technology, can improve the testing efficiency and accuracy of photovoltaic products. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 A flowchart of a photovoltaic product testing method provided by an embodiment of the present invention is shown; Figure 2 A flowchart of another photovoltaic product testing method provided by an embodiment of the present invention is shown; Figure 3 This diagram illustrates the structure of a photovoltaic product testing device according to an embodiment of the present invention. Figure 4 This invention provides a schematic diagram of the structure of another photovoltaic product testing device according to an embodiment of the invention. Figure 5 A schematic diagram of the physical structure of a computer device provided in an embodiment of the present invention is shown. Detailed Implementation

[0017] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0018] Currently, manually inspecting photovoltaic products is time-consuming and labor-intensive, and due to the negligence of staff and the varying levels of technical expertise, it can lead to missed or incorrect inspections.

[0019] To address the above problems, embodiments of the present invention provide a method for testing photovoltaic products, such as... Figure 1 As shown, the method includes: 101. Real-time acquisition of multiple RGBD images of the photovoltaic product under test as it moves along the production line.

[0020] The photovoltaic products to be tested can include photovoltaic panels, photovoltaic cells, and mounting brackets.

[0021] In this embodiment of the invention, RGBD cameras are positioned on both sides of the photovoltaic product production line, i.e., multiple RGBD cameras are staggered along the length of the production line. Each RGBD camera acquires RGBD images of the photovoltaic products to be inspected as they pass through the production line in real time. All acquired RGBD images are uploaded to a cloud server, and then defects or quality are detected in the photovoltaic products based on these RGBD images. This embodiment of the invention uses visual technology to inspect photovoltaic products, which can improve inspection efficiency and quality.

[0022] 102. Perform image decomposition on each RGBD image, extract image contours from the decomposed RGBD images, and determine the line drawing of the photovoltaic product to be tested based on the extracted image contours.

[0023] In this embodiment of the invention, the RGBD image stored on the server cloud is decomposed to remove color spots, and the image outline is extracted. The image outline is then decomposed into line graphs containing line segments, polylines, circles, arcs, ellipses, and text. This embodiment of the invention uses line graphs to detect photovoltaic products, removing redundant information such as color and texture from the image. This reduces computational resource consumption and improves the detection efficiency of photovoltaic products.

[0024] 103. Match the line drawing with the standard line drawing in the preset photovoltaic product library. Based on the matching result, determine the test result of the photovoltaic product to be tested. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

[0025] In this invention, a qualified photovoltaic product refers to a photovoltaic product that is of acceptable quality and free of defects. In this embodiment, a line drawing is sent to the client, where it is matched against standard line drawings in a pre-built photovoltaic product library. If a perfectly matching standard line drawing exists, the photovoltaic product is deemed qualified; otherwise, it is deemed unqualified. This invention determines the inspection result of a photovoltaic product by extracting a line drawing from an RGBD image and matching it with standard line drawings in a pre-built photovoltaic product library. This avoids the time wasted on manual inspection and the risk of errors. Therefore, this invention, through visual inspection technology, improves the inspection efficiency and accuracy of photovoltaic products.

[0026] In another embodiment of the present invention, it is necessary to update the preset photovoltaic product database periodically to meet the testing requirements of subsequent photovoltaic products.

[0027] According to the present invention, a method for testing photovoltaic products, compared with the current method of manually inspecting photovoltaic products, involves acquiring multiple RGBD images of the photovoltaic product to be tested as it moves along the production line in real time. Then, each RGBD image is decomposed, and image contours are extracted from the decomposed RGBD images. Based on the extracted image contours, a line drawing of the photovoltaic product to be tested is determined. Finally, the line drawing is matched with standard line drawings in a preset photovoltaic product library. Based on the matching result, the test result of the photovoltaic product to be tested is determined. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products. Therefore, by extracting the line drawing of the photovoltaic product from the RGBD images and matching it with the standard line drawings in the preset photovoltaic product library to determine the test result, the time wasted and errors associated with manual inspection can be avoided. Thus, the present invention, through visual inspection technology, can improve the testing efficiency and accuracy of photovoltaic products.

[0028] Furthermore, to better illustrate the above-described process for testing photovoltaic products, and as a refinement and extension of the above embodiments, this invention provides another method for testing photovoltaic products, such as... Figure 2 As shown, the method includes: 201. Real-time acquisition of multiple RGBD images of the photovoltaic product under test as it moves along the production line.

[0029] Specifically, the photovoltaic products to be tested are photographed in real time by cameras that are staggered on the left and right sides of the production line, resulting in multiple RGBD images captured by the left camera and multiple RGBD images captured by the right camera.

[0030] 202. Perform image decomposition on each RGBD image and extract image contours from the decomposed RGBD images. Based on the extracted image contours, determine the line graph of the photovoltaic product to be tested. The line graph is the line graph corresponding to the RGBD images captured by multiple cameras arranged alternately on both sides of the production line. It includes the left line graph corresponding to the multiple left RGBD images captured by the cameras arranged on the left side of the production line, and the right line graph corresponding to the multiple right RGBD images captured by the cameras arranged on the right side of the production line.

[0031] In this embodiment of the invention, in order to extract line graphs from an RGBD image, the RGBD image first needs to be decomposed. Based on this, the method includes: for each RGBD image, performing RGB channel separation to obtain multiple RGB channel images; based on the pixel value distribution information of each RGB channel image, determining color spot regions in the corresponding RGB channel image, and removing the color spot regions in the corresponding RGB channel image to obtain each de-spotted RGB channel image, and the de-spotted RGB channel images constitute the separated RGBD image.

[0032] Pixel value distribution information refers to the magnitude of pixel values ​​in each RGB channel image. Specifically, the red, green, and blue channels of the RGBD image are extracted independently. For example, an RGB image of a photovoltaic panel can be decomposed into: Red channel: highlighting the reflective areas of the metal grid lines (e.g., silver-white grid lines are more prominent in the red channel); Green channel: displaying substrate uniformity (e.g., the silicon wafer substrate exhibits uniform grayscale in the green channel); Blue channel: capturing edge details (e.g., the photovoltaic panel frame has the highest contrast in the blue channel). Furthermore, by statistically analyzing the pixel value distribution of each channel image, abnormally bright / dark areas are identified as color spots. For example, yellow spots on the surface of photovoltaic products caused by oil stains appear in the RGB channels as local RGB values ​​deviating from the normal range (e.g., a normal silicon wafer has R=120±10, while the color spot area has R=180). Adaptive median filtering combined with morphological closure operations is used to process the color spot areas in the RGB channel images. For example, for yellow spots, isolated noise is first eliminated by 3×3 median filtering, and then holes are filled by a morphological "dilation-erosion" combination operation, making the spots blend smoothly with the surrounding substrate. After processing, the originally abrupt spots are "softened" to a grayscale value similar to the substrate, thereby improving the accuracy of subsequent contour extraction. The resulting RGBD images are obtained in this way. Next, image contour extraction and line drawing extraction are required for the separated RGBD images. Based on this, the method includes: determining the depth map of the decomposed RGBD image; extracting the image edges of the decomposed RGBD image and determining whether there are elevation changes in the depth map; if so, the image edges are determined as the image contour of the decomposed RGBD image; otherwise, the image edges are determined as pseudo-contours and filtered; decomposing the image contour into multiple vector graphic primitives, wherein the multiple vector graphic primitives include at least one of line segments, polylines, circles, arcs, ellipses, and text; combining each vector graphic primitive according to its original position in the image contour, and obtaining the line drawing of the photovoltaic product to be detected based on the combination result.

[0033] Specifically, the depth value of each pixel in the decomposed RGBD image is obtained, and a depth map is formed based on the depth values. The Canny operator can be applied to the RGBD image to detect image edges such as the outer frame and grid line intersections of the photovoltaic product. If an image edge exhibits a corresponding elevation change in the depth map (e.g., the height difference between the frame and the base > 0.5mm), the image edge is confirmed as a valid image contour; if the depth change is gradual (e.g., a gradient shadow), the image edge is determined to be a false edge and filtered out. The extracted image contours are then subjected to "skeletonization." For example, grid line edges with a width of 3-5 pixels are thinned to 1 pixel to ensure the accuracy of subsequent geometric decomposition. Furthermore, straight line segments in the image contours are detected using methods such as Hough transform. For example, the four long sides of the photovoltaic panel frame can be identified as four straight line segments, and the grid lines are decomposed into multiple short straight lines or broken lines (e.g., curved welding strips are fitted as broken lines). The least squares method is used to fit the arcs in the image contour. For example, the chamfered arcs at the four corners of the photovoltaic panel can be accurately fitted as arcs with a radius R=5mm, while the slightly curved parts at the edges are fitted as elliptical arcs. Text regions in the image contour are extracted using techniques such as OCR. For example, the model designation "PV-250W" on the photovoltaic panel is identified as a text element, and its contour is extracted and vectorized separately. Non-circular but elliptical shapes are fitted; for example, the edges of some specially designed photovoltaic panels may have an elliptical shape, and their major and minor axis parameters are accurately extracted using an ellipse fitting algorithm. Thus, various vector graphic primitives such as line segments, polylines, circles, arcs, ellipses, and text can be extracted from the image contour in the above manner. Furthermore, the identified line segments, polylines, circles, arcs, ellipses, and text elements are combined according to their positions in the original image contour to form vectorized line drawings. For example, the outer frame of the photovoltaic panel consists of four straight line segments and four arcs, the grid lines are represented by multiple polylines, and the model text is individually labeled in a designated position.

[0034] 203. Take the RGBD image on either the left or right side of the RGBD image as the target RGBD image, determine multiple current frame feature points in the current frame RGBD image of the target RGBD image, and determine multiple previous frame feature points in the previous frame RGBD image on the same side as the current frame RGBD image.

[0035] Specifically, taking the left-side RGBD image captured on the left side of the production line as an example, points with significant features, such as corner points and grid line intersections of the photovoltaic product to be detected, are selected as feature points in the current frame of the left-side RGBD image. Similarly, points with significant features, such as corner points and grid line intersections of the photovoltaic product to be detected, are selected as feature points in the previous frame of the left-side RGBD image.

[0036] 204. Perform feature point matching between the feature points of the current frame and the feature points of the previous frame to obtain successfully matched feature point pairs. Based on the position information of the feature point pairs, determine the pose transformation parameters of the photovoltaic product to be detected along the production line.

[0037] Specifically, feature point pairs are formed by identifying points at the same location on the photovoltaic product to be detected from the feature points of the current frame and the feature points of the previous frame. Based on the 3D coordinate differences of the feature point pairs, pose transformation parameters such as the movement vector (translation T) and rotation angle (rotation R) of the photovoltaic product to be detected are calculated. For example, the translation T is [X=5mm, Y=0, Z=0], and the rotation R is a rotation of 2 degrees around the Z-axis, forming an RT matrix, which indicates that the photovoltaic product to be detected has translated 5mm to the right and rotated slightly from the previous frame to the current frame.

[0038] 205. Determine the 3D point cloud of the previous frame of the line graph corresponding to the previous frame of the RGBD image and the 3D point cloud of the current frame of the line graph corresponding to the current frame of the RGBD image.

[0039] Specifically, the line graph corresponding to the previous RGBD image is converted into a 3D point cloud P0, and the line graph corresponding to the current RGBD image is converted into a 3D point cloud P1. Each point in the 3D point cloud represents a pixel on the line graph, and the depth information of the point cloud is determined through the depth channel of the RGBD image.

[0040] 206. Based on the pose transformation parameters, the 3D point cloud of the previous frame is transformed to the current coordinate system of the 3D point cloud of the current frame, and the transformed 3D point cloud of the previous frame is fused with the 3D point cloud of the current frame to obtain the fused 3D point cloud.

[0041] Specifically, the 3D point cloud P0 generated from the line graph corresponding to the previous RGBD image is transformed to the coordinate system of the current RGBD image using a Real-Time Transform (RT) matrix. For example, the top-left corner point (0,0,0) in the previous 3D point cloud P0 becomes (5,0,0) after RT transformation, aligning with the corresponding point in the current point cloud P1. Then, the transformed P0 is fused with the current frame P1 to generate a fused 3D point cloud P2. For example, the border points of P0 are superimposed with the border points of P1, forming a denser point cloud and reducing holes caused by single-frame noise.

[0042] 207. Determine the degree of matching between the fused 3D point cloud and the current frame's 3D point cloud. Based on the degree of matching, select the target line map for subsequent matching from the left line map corresponding to the left RGBD image and the right line map corresponding to the right RGBD image.

[0043] In this embodiment of the invention, in order to select a target line drawing that can be matched subsequently, it is first necessary to determine the degree of matching between the fused 3D point cloud and the current frame 3D point cloud. Based on this, step 207 specifically includes: calculating the distance between each fused point in the fused 3D point cloud and each current point in the current frame 3D point cloud, and determining the minimum distance corresponding to each fused point in each distance, and determining the maximum distance as the maximum fused distance in each minimum distance; calculating the distance between each current point in the current frame 3D point cloud and each fused point in the fused 3D point cloud, and determining the minimum distance corresponding to each current point in each distance, and determining the maximum distance as the maximum current distance in each minimum distance; based on the maximum fused distance and the maximum current distance... The following steps are taken: First, determine the global shape matching degree between the fused 3D point cloud and the current frame 3D point cloud. Second, determine multiple curvature intervals based on the curvature values ​​of each point in the fused 3D point cloud and the current frame 3D point cloud. Third, determine the fusion probability density of the fused 3D point cloud and the current probability density of the current frame 3D point cloud in each curvature interval, respectively. Then, determine the fusion curvature evaluation value of the fused 3D point cloud based on the fusion probability density, and determine the current curvature evaluation value of the current frame 3D point cloud based on the current probability density. Fourth, determine the local edge matching degree between the fused 3D point cloud and the current frame 3D point cloud based on the fusion curvature evaluation value and the current curvature evaluation value. Finally, determine the degree of matching between the fused 3D point cloud and the current frame 3D point cloud based on the global shape matching degree and the local edge matching degree.

[0044] Specifically, for example, if the 3D point cloud is fused as The current frame's 3D point cloud is For each point in the fused 3D point cloud, calculate its distance to all points in the current frame's 3D point cloud and take the minimum value. For example, the distances from point (1,2) to each point in the current frame's 3D point cloud are shown below: The distance to (2,3) is: ; The distance to (4,5) is: ; The distance to (6,7) is: .

[0045] From the above distances, we can see that the minimum distance is... Repeat the above steps for the remaining points in the fused 3D point cloud to obtain a set of minimum distances. Take the maximum value of this set of minimum distances as the maximum fusion distance. Similarly, the maximum current distance is calculated in the same way as described above. This involves calculating the minimum distance from each point in the current frame's 3D point cloud to all points in the fused 3D point cloud, and then taking the maximum of these minimum distances. The maximum distance is then selected from the maximum fused distance and the maximum current distance as the evaluation distance, and this evaluation distance is normalized to obtain the global shape matching degree. .

[0046] Furthermore, if the curvature values ​​of each point in the fused 3D point cloud and the current frame 3D point cloud are calculated, and these curvature values ​​are divided into several intervals (e.g., three intervals: interval 1, interval 2, and interval 3), the fusion probability density of the fused 3D point cloud and the current probability density of the current frame 3D point cloud in each interval are determined based on the number of points in each interval. Then, the fusion curvature evaluation value and the current curvature evaluation value are determined according to the following formula:

[0047] in, To integrate curvature evaluation values ​​( ) or the current curvature evaluation value ( ), To fuse the probability density of the 3D point cloud or the current frame's 3D point cloud in the i-th curvature interval, N is the total number of curvature intervals. Then, the local edge matching degree is calculated according to the following formula. :

[0048] Furthermore, the weight coefficients corresponding to the global shape matching degree and the local edge matching degree are determined according to actual needs. Based on the weight coefficients, the global shape matching degree and the local edge matching degree are weighted and summed. Finally, the weighted summation result is used as the degree of matching between the fused 3D point cloud and the current frame 3D point cloud.

[0049] Furthermore, it is also necessary to select a target line graph based on the matching degree. Based on this, step 207 specifically includes: acquiring motion feature data of the photovoltaic product to be detected between the previous frame RGBD image and the current frame RGBD image, wherein the motion feature data includes motion speed and vibration frequency; dynamically determining a preset matching degree threshold based on the motion feature data; determining whether the matching degree is greater than the preset matching degree threshold; if so, selecting any line graph from the line graphs corresponding to the target side RGBD image as the target line graph; otherwise, selecting any line graph from the line graphs corresponding to the other side RGBD image corresponding to the target side RGBD image as the target line graph.

[0050] Specifically, the preset matching degree threshold is calculated according to the following formula. :

[0051] in, This is the initial matching threshold. The speed weighting coefficient is set according to actual needs. The vibration weighting coefficient is set according to actual needs. For the speed of movement, The vibration frequency is used. Specifically, if the matching degree is greater than a preset matching degree threshold, a line graph is randomly selected from the line graphs corresponding to the target RGBD image (such as the left RGBD image) as the target line graph; otherwise, a line graph is randomly selected from the right RGBD image as the target line graph. This embodiment of the invention, by dynamically setting the threshold, can solve the problem of misjudgment in high-speed or vibration scenarios using traditional fixed thresholds, thereby improving the detection accuracy of photovoltaic products. By selecting a suitable target line graph for photovoltaic product detection, high-quality, information-effective line graphs can be quickly screened for subsequent analysis, avoiding the transmission of a large number of poor-quality or duplicate line graphs to subsequent processing stages, reducing data transmission and processing time, and further improving the efficiency of the overall detection process.

[0052] 208. Match the target line drawing with the standard line drawing in the preset photovoltaic product library. Based on the matching results, determine the test result of the photovoltaic product to be tested. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

[0053] In this embodiment of the invention, after determining the target line drawing for matching, it is necessary to match the target line drawing with a standard line drawing in a preset photovoltaic product library. Based on this, step 208 specifically includes: dividing the outer contour of the standard line drawing into a preset number of standard points, and determining a detection point corresponding to each standard point on the line drawing; determining the point distance between each standard point and its corresponding detection point; determining the angle between two adjacent standard points in each standard point, and determining the angle between two adjacent detection points corresponding to the two adjacent standard points; determining the standard extension direction angle corresponding to each standard point based on the positional relationship between each standard point and its corresponding adjacent standard points; and determining the standard extension direction angle between each detection point and its corresponding adjacent standard points. Based on the positional relationship between adjacent points to be detected, determine the angle of the extension direction to be detected corresponding to each point to be detected; based on the point distance, perform distance matching between the line drawing and the standard line drawing in the preset photovoltaic product library to obtain a distance matching result; based on the included angle of the point and the included angle of the point to be detected, perform included angle matching between the line drawing and the standard line drawing in the preset photovoltaic product library to obtain an included angle matching result; based on the standard extension direction angle and the included extension direction angle to be detected, perform extension angle matching between the line drawing and the standard line drawing in the preset photovoltaic product library to obtain an extension angle matching result; based on the distance matching result, the included angle matching result, and the extension angle matching result, determine the matching result between the line drawing and the standard line drawing in the preset photovoltaic product library.

[0054] The preset quantity is set according to actual needs, such as 4. Specifically, on the outer contour of the standard line drawing, four points are evenly divided according to equal angles or equal arc lengths, denoted as points A, B, C, and D. For example, assuming the outer contour of the photovoltaic panel is an approximate rectangle, we can use the four vertices of the rectangle as these four evenly divided points; if the outer contour is an irregular shape, four representative points are selected on the contour according to a certain algorithm (such as equal angle intervals). Similarly, on the outer contour of the target line drawing, find the positions corresponding to the four points on the standard line drawing, and mark four points as points A', B', C', and D'. Calculate the distance between point A on the standard line drawing and the corresponding point A' on the target line drawing, and then calculate the distances between points B and B', C and C', and D and D' in turn. Set a distance threshold. If all four distances are less than the distance threshold, it means that the target line drawing and the standard line drawing are relatively close in terms of point position distance. If any distance exceeds the distance threshold, it may indicate that there is a difference in shape between the two, and the shape is initially judged to be dissimilar, i.e., the test is unqualified. Meanwhile, on the standard line drawing, the angle between two adjacent points is calculated using vectors. On the target line drawing, the angle between corresponding adjacent points is calculated, and an angle threshold is set. The angles on the standard and target line drawings are compared. If the difference between all corresponding angles is less than the angle threshold, the two are similar in terms of angle. If any angle difference exceeds the angle threshold, the shapes may differ, indicating that the test result is unqualified. Simultaneously, for each point on the standard line drawing, its extension direction angle is determined. For example, at point A, its extension direction angle can be determined based on the positional relationship of adjacent points. This can be achieved by calculating the slope of the line connecting point A and its adjacent points and then converting the slope into an angle. Similarly, the extension direction angle of the corresponding point A' on the target line drawing is calculated, and an extension direction angle threshold is set. The extension direction angles of corresponding points on the standard and target line drawings are compared. If the difference between all corresponding point extension direction angles is less than the extension direction angle threshold, the two are similar in terms of extension direction. If any difference exceeds the extension direction angle threshold, the shapes may be dissimilar. Finally, based on the comparison results of the comprehensive distance, included angle, and extension direction angle, if all indicators are within the corresponding threshold range, the target line diagram is judged to be similar in shape to the standard line diagram, and the photovoltaic product to be tested is a qualified product; if any indicator exceeds the threshold range, the shape is judged to be dissimilar, and the photovoltaic product to be tested is a non-qualified product.

[0055] According to another photovoltaic product testing method provided by the present invention, compared with the current method of manually inspecting photovoltaic products, the present invention acquires multiple RGBD images of the photovoltaic product to be tested as it moves along the production line in real time; then, it decomposes each RGBD image and extracts the image contour from the decomposed RGBD images, determining the line drawing of the photovoltaic product to be tested based on the extracted image contour; finally, it matches the line drawing with standard line drawings in a preset photovoltaic product library, and determines the test result of the photovoltaic product to be tested based on the matching result. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products. Therefore, by extracting the line drawing of the photovoltaic product from the RGBD images and matching it with the standard line drawings in the preset photovoltaic product library to determine the test result, the time wasted by manual inspection and the possibility of inspection errors can be avoided. Thus, the present invention, through visual inspection technology, can improve the inspection efficiency and accuracy of photovoltaic products.

[0056] Furthermore, as Figure 1 In specific implementation, embodiments of the present invention provide a testing device for photovoltaic products, such as... Figure 3 As shown, the device includes: an acquisition unit 31, a determination unit 32, and a matching unit 33.

[0057] The acquisition unit 31 can be used to acquire multiple RGBD images of the photovoltaic product under test as it moves along the production line in real time.

[0058] The determining unit 32 can be used to decompose each of the RGBD images, extract the image contours of the decomposed RGBD images, and determine the line drawing of the photovoltaic product to be detected based on the extracted image contours.

[0059] The matching unit 33 can be used to match the line drawing with the standard line drawing in the preset photovoltaic product library, and determine the test result of the photovoltaic product to be tested based on the matching result. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

[0060] In specific application scenarios, the line graph refers to the line graph corresponding to RGBD images captured by multiple camera devices staggered on both sides of the production line. This includes left-side line graphs corresponding to multiple left-side RGBD images captured by cameras positioned on the left side of the production line, and right-side line graphs corresponding to multiple right-side RGBD images captured by cameras positioned on the right side of the production line. To determine the target line graph, such as... Figure 4 As shown, the device also includes a selection unit 34.

[0061] The selection unit 34 can be used to select either the left RGBD image or the right RGBD image as the target RGBD image, determine multiple current frame feature points in the current frame RGBD image of the target RGBD image, and determine multiple previous frame feature points in the previous frame RGBD image on the same side as the current frame RGBD image; perform feature point matching on the current frame feature points and the previous frame feature points to obtain successfully matched feature point pairs, and determine the pose transformation parameters of the photovoltaic product to be detected along the production line based on the position information of the feature point pairs; and determine the previous frame RGBD image. The previous frame 3D point cloud of the line graph corresponding to the GBD image and the current frame 3D point cloud of the line graph corresponding to the current frame RGBD image are used. Based on the pose transformation parameters, the previous frame 3D point cloud is transformed to the current coordinate system of the current frame 3D point cloud, and the transformed previous frame 3D point cloud and the current frame 3D point cloud are fused to obtain a fused 3D point cloud. The matching degree between the fused 3D point cloud and the current frame 3D point cloud is determined, and based on the matching degree, a target line graph for subsequent matching is selected from the left line graph corresponding to the left RGBD image and the right line graph corresponding to the right RGBD image.

[0062] In specific application scenarios, in order to match the line drawing with the standard line drawing in the preset photovoltaic product library, the matching unit 33 can be used to match the target line drawing with the standard line drawing in the preset photovoltaic product library.

[0063] In specific application scenarios, in order to determine the degree of matching between the fused 3D point cloud and the current frame 3D point cloud, the selection unit 34 includes a calculation module 341 and a first determination module 342.

[0064] The calculation module 341 can be used to calculate the distance between each fusion point in the fused 3D point cloud and each current point in the current frame 3D point cloud, and determine the minimum distance corresponding to each fusion point in each distance, and determine the maximum distance as the maximum fusion distance in the minimum distance corresponding to each fusion point.

[0065] The calculation module 341 can also be used to calculate the distance between each current point in the current frame 3D point cloud and each fusion point in the fused 3D point cloud, and determine the minimum distance corresponding to each current point in each distance, and determine the maximum distance as the maximum current distance in each minimum distance.

[0066] The first determining module 342 can be used to determine the global shape matching degree between the fused 3D point cloud and the current frame 3D point cloud based on the maximum fusion distance and the maximum current distance.

[0067] The first determining module 342 can also be used to determine multiple curvature intervals based on the curvature values ​​of each point in the fused 3D point cloud and the curvature values ​​of each point in the current frame 3D point cloud.

[0068] The first determining module 342 can be specifically used to determine the fusion probability density of the fused 3D point cloud in each curvature interval and the current probability density of the current frame 3D point cloud in each curvature interval, and determine the fusion curvature evaluation value of the fused 3D point cloud based on the fusion probability density, and determine the current curvature evaluation value of the current frame 3D point cloud based on the current probability density.

[0069] The first determining module 342 can be specifically used to determine the local edge matching degree between the fused 3D point cloud and the current frame 3D point cloud based on the fused curvature evaluation value and the current curvature evaluation value.

[0070] The first determining module 342 can be specifically used to determine the degree of matching between the fused 3D point cloud and the current frame 3D point cloud based on the global shape matching degree and the local edge matching degree.

[0071] In specific application scenarios, in order to select a target line drawing, the selection unit 34 further includes an acquisition module 343 and a judgment module 344.

[0072] The acquisition module 343 is used to acquire motion feature data of the photovoltaic product to be detected between the previous frame RGBD image and the current frame RGBD image, wherein the motion feature data includes motion speed and vibration frequency.

[0073] The first determining module 342 can also be used to dynamically determine a preset matching degree threshold based on the motion feature data.

[0074] The judgment module 344 can be used to determine whether the matching degree is greater than the preset matching degree threshold. If so, it can select any line drawing in the line drawing corresponding to the target RGBD image as the target line drawing. Otherwise, it can select any line drawing in the line drawing corresponding to the other RGBD image corresponding to the target RGBD image as the target line drawing.

[0075] In specific application scenarios, in order to match the line drawing with the standard line drawing in the preset photovoltaic product library, the matching unit 33 includes an equal division module 331, a second determination module 332, and a matching module 333.

[0076] The equalization module 331 can be used to evenly divide a preset number of standard points on the outer contour of the standard line drawing, and to determine the detection point corresponding to each standard point on the line drawing.

[0077] The second determining module 332 can be used to determine the point distance between each of the standard points and the corresponding point to be detected.

[0078] The second determining module 332 can also be used to determine the angle between two adjacent standard points in each of the standard points, and to determine the angle between two adjacent test points corresponding to the two adjacent standard points.

[0079] The second determining module 332 can also be used to determine the standard extension direction angle corresponding to each standard point based on the positional relationship between each standard point and its corresponding adjacent standard point, and to determine the detection extension direction angle corresponding to each detection point based on the positional relationship between each detection point and its corresponding adjacent detection point.

[0080] The matching module 333 can be used to perform distance matching between the line drawing and the standard line drawing in the preset photovoltaic product library based on the point distance to obtain a distance matching result; to perform angle matching between the line drawing and the standard line drawing in the preset photovoltaic product library based on the point angle and the angle of the point to be detected to obtain an angle matching result; and to perform extension angle matching between the line drawing and the standard line drawing in the preset photovoltaic product library based on the standard extension direction angle and the extension direction angle to be detected to obtain an extension angle matching result.

[0081] The second determining module 332 can be specifically used to determine the matching result between the line drawing and the standard line drawing in the preset photovoltaic product library based on the distance matching result, the included angle matching result, and the extension angle matching result.

[0082] In specific application scenarios, in order to extract the line drawing of the photovoltaic product to be tested, the determining unit 32 includes a third determining module 321, an extraction module 322, a decomposition module 323, and a combination module 324.

[0083] The third determining module 321 can be used to determine the depth map of the decomposed RGBD image.

[0084] The extraction module 322 can be used to extract the image edges of the decomposed RGBD image and determine whether there are elevation changes in the depth map. If so, the image edges are determined as the image contours of the decomposed RGBD image; otherwise, the image edges are determined as pseudo-contours and filtered.

[0085] The decomposition module 323 can be used to decompose the image outline into a variety of vector graphic primitives, wherein the variety of vector graphic primitives includes at least one of line segments, polylines, circles, arcs, ellipses, and text.

[0086] The combination module 324 can be used to combine each of the vector graphic primitives according to their original positions in the image outline, and based on the combination result, obtain the line drawing of the photovoltaic product to be tested.

[0087] In specific application scenarios, in order to decompose each RGBD image separately, the decomposition module 323 can also be used to separate the RGB channels of each RGBD image to obtain multiple RGB channel images; based on the pixel value distribution information of each RGB channel image, the color spot region is determined in the corresponding RGB channel image, and the color spot region is removed in the corresponding RGB channel image to obtain each despotted RGB channel image, and the despotted RGB channel images constitute the separated RGBD images.

[0088] It should be noted that other corresponding descriptions of the functional modules involved in the photovoltaic product testing device provided in this embodiment of the invention can be found in [reference]. Figure 1 The corresponding description of the method shown will not be repeated here.

[0089] Based on the above, Figure 1 Accordingly, this embodiment of the invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the following steps: acquiring multiple RGBD images of the photovoltaic product to be tested moving along the production line in real time; performing image decomposition on each RGBD image and extracting image contours from the decomposed RGBD images; determining the line drawing of the photovoltaic product to be tested based on the extracted image contours; matching the line drawing with standard line drawings in a preset photovoltaic product library; and determining the test result of the photovoltaic product to be tested based on the matching result, wherein the preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

[0090] Based on the above, Figure 1 The method shown and as Figure 3 The embodiment of the device shown in the invention also provides a physical structure diagram of a computer device, such as... Figure 5As shown, the computer device includes: a processor 41, a memory 42, and a computer program stored in the memory 42 and executable on the processor. Both the memory 42 and the processor 41 are mounted on a bus 43. When the processor 41 executes the program, it performs the following steps: real-time acquisition of multiple RGBD images of the photovoltaic product to be tested moving along the production line; image decomposition of each RGBD image, and image contour extraction of the decomposed RGBD images; determining the line drawing of the photovoltaic product to be tested based on the extracted image contours; matching the line drawing with standard line drawings in a preset photovoltaic product library; and determining the detection result of the photovoltaic product to be tested based on the matching result. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

[0091] The present invention acquires multiple RGBD images of a photovoltaic product under inspection as it moves along a production line in real time. Each RGBD image is then decomposed, and image contours are extracted from the decomposed images. Based on the extracted image contours, a line drawing of the photovoltaic product under inspection is determined. Finally, the line drawing is matched with standard line drawings in a preset photovoltaic product library. Based on the matching result, the inspection result of the photovoltaic product is determined. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products. Therefore, by extracting the line drawing of the photovoltaic product from the RGBD images and matching it with standard line drawings in the preset photovoltaic product library to determine the inspection result, the time wasted on manual inspection and the possibility of errors can be avoided. Thus, the present invention, through visual inspection technology, can improve the inspection efficiency and accuracy of photovoltaic products.

[0092] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0093] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for testing photovoltaic products, characterized in that, include: Multiple RGBD images of the photovoltaic product under test are acquired in real time as it moves along the production line. Each RGBD image is decomposed, and the decomposed RGBD images are used to extract image contours. Based on the extracted image contours, the line drawing of the photovoltaic product to be detected is determined. The line drawing is matched with the standard line drawing in the preset photovoltaic product library. Based on the matching result, the test result of the photovoltaic product to be tested is determined. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

2. The method according to claim 1, characterized in that, The line chart is a line chart corresponding to RGBD images captured by multiple camera devices arranged alternately on both sides of the production line, including left-side line charts corresponding to multiple left-side RGBD images captured by camera devices arranged on the left side of the production line, and right-side line charts corresponding to multiple right-side RGBD images captured by camera devices arranged on the right side of the production line. Before matching the line drawing with a standard line drawing in a preset photovoltaic product library, the method further includes: Take either the left RGBD image or the right RGBD image as the target RGBD image, determine multiple current frame feature points in the current frame RGBD image of the target RGBD image, and determine multiple previous frame feature points in the previous frame RGBD image on the same side as the current frame RGBD image. Feature point matching is performed on the feature points of the current frame and the feature points of the previous frame to obtain successfully matched feature point pairs. Based on the position information of the feature point pairs, the pose transformation parameters of the photovoltaic product to be detected along the production line are determined. The three-dimensional point cloud of the previous frame of the line graph corresponding to the previous frame RGBD image and the three-dimensional point cloud of the current frame of the line graph corresponding to the current frame RGBD image are determined respectively. Based on the pose transformation parameters, the previous frame 3D point cloud is transformed into the current coordinate system where the current frame 3D point cloud is located, and the transformed previous frame 3D point cloud and the current frame 3D point cloud are fused to obtain a fused 3D point cloud. Determine the degree of matching between the fused 3D point cloud and the current frame 3D point cloud, and based on the degree of matching, select a target line image for subsequent matching from the left line image corresponding to the left RGBD image and the right line image corresponding to the right RGBD image. The step of matching the line drawing with a standard line drawing in a preset photovoltaic product library includes: The target line drawing is matched with the standard line drawing in the preset photovoltaic product library.

3. The method according to claim 2, characterized in that, Determining the degree of matching between the fused 3D point cloud and the current frame 3D point cloud includes: Calculate the distance between each fusion point in the fused 3D point cloud and each current point in the current frame 3D point cloud, determine the minimum distance corresponding to each fusion point in each distance, and determine the maximum distance in the minimum distance corresponding to each fusion point as the maximum fusion distance; Calculate the distance between each current point in the current frame 3D point cloud and each fused point in the fused 3D point cloud, and determine the minimum distance corresponding to each current point in each distance, and determine the maximum distance in each minimum distance as the maximum current distance; Based on the maximum fusion distance and the maximum current distance, determine the global shape matching degree between the fused 3D point cloud and the current frame 3D point cloud; Based on the curvature values ​​of each point in the fused 3D point cloud and the curvature values ​​of each point in the current frame 3D point cloud, multiple curvature intervals are determined; The fusion probability density of the fused 3D point cloud in each curvature interval and the current probability density of the current frame 3D point cloud in each curvature interval are determined respectively. The fusion curvature evaluation value of the fused 3D point cloud is determined based on the fusion probability density, and the current curvature evaluation value of the current frame 3D point cloud is determined based on the current probability density. Based on the fused curvature evaluation value and the current curvature evaluation value, the local edge matching degree between the fused 3D point cloud and the current frame 3D point cloud is determined; Based on the global shape matching degree and the local edge matching degree, the degree of matching between the fused 3D point cloud and the current frame 3D point cloud is determined.

4. The method according to claim 2, characterized in that, Based on the matching degree, selecting the target line image for subsequent matching from the left line image corresponding to the left RGBD image and the right line image corresponding to the right RGBD image includes: The motion feature data of the photovoltaic product to be detected between the previous frame RGBD image and the current frame RGBD image is obtained, wherein the motion feature data includes motion speed and vibration frequency; Based on the motion feature data, a preset matching degree threshold is dynamically determined; If the matching degree is greater than the preset matching degree threshold, then select any line drawing from the line drawing corresponding to the target RGBD image as the target line drawing; otherwise, select any line drawing from the line drawing corresponding to the other RGBD image corresponding to the target RGBD image as the target line drawing.

5. The method according to claim 1, characterized in that, The step of matching the line drawing with a standard line drawing in a preset photovoltaic product library includes: A predetermined number of standard points are evenly distributed on the outer contour of the standard line drawing, and a detection point corresponding to each standard point is determined on the line drawing; Determine the point distance between each of the aforementioned standard points and the corresponding point to be detected; Determine the angle between two adjacent standard points in each of the standard points, and determine the angle between two adjacent test points corresponding to the two adjacent standard points; Based on the positional relationship between each standard point and its corresponding adjacent standard points, the standard extension direction angle corresponding to each standard point is determined; based on the positional relationship between each point to be detected and its corresponding adjacent points to be detected, the detection extension direction angle corresponding to each point to be detected is determined. Based on the point distance, the line drawing is matched with the standard line drawing in the preset photovoltaic product library to obtain a distance matching result. Based on the point angle and the angle of the point to be detected, the line drawing is matched with the standard line drawing in the preset photovoltaic product library to obtain an angle matching result. Based on the standard extension direction angle and the extension direction angle to be detected, the line drawing is matched with the standard line drawing in the preset photovoltaic product library to obtain an extension angle matching result. Based on the distance matching result, the included angle matching result, and the extension angle matching result, the matching result between the line drawing and the standard line drawing in the preset photovoltaic product library is determined.

6. The method according to claim 1, characterized in that, The step of extracting image contours from the decomposed RGBD image and determining the line drawing of the photovoltaic product to be detected based on the extracted image contours includes: Determine the depth map of the decomposed RGBD image; Extract the image edges of the decomposed RGBD image and determine whether there are elevation changes in the depth map. If so, the image edges are determined as the image contours of the decomposed RGBD image; otherwise, the image edges are determined as pseudo contours and filtered out. The image outline is decomposed into a variety of vector graphic primitives, wherein the variety of vector graphic primitives includes at least one of line segments, polylines, circles, arcs, ellipses, and text; Each of the vector graphic primitives is combined according to its original position in the image outline, and based on the combination result, a line drawing of the photovoltaic product to be tested is obtained.

7. The method according to claim 1, characterized in that, Each of the RGBD images is decomposed, including: For each RGBD image, the RGB channels are separated to obtain multiple RGB channel images; Based on the pixel value distribution information of each RGB channel image, a color spot region is determined in the corresponding RGB channel image, and the color spot region is removed in the corresponding RGB channel image to obtain each despotted RGB channel image. Each despotted RGB channel image constitutes the RGBD image.

8. A testing device for photovoltaic products, characterized in that, include: The acquisition unit is used to acquire multiple RGBD images of the photovoltaic product under test as it moves along the production line in real time. The determining unit is used to decompose each of the RGBD images, extract the image contours of the decomposed RGBD images, and determine the line drawing of the photovoltaic product to be detected based on the extracted image contours. The matching unit is used to match the line drawing with the standard line drawing in the preset photovoltaic product library, and determine the test result of the photovoltaic product to be tested based on the matching result. The preset photovoltaic product library stores standard line drawings corresponding to various qualified photovoltaic products.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images