Connecting piece welding quality detection method and system, medium and electronic equipment

Abstract

The invention discloses a connecting piece welding quality detection method and system, a medium and electronic equipment, and relates to the technical field of industrial automation and machine vision. The method specifically comprises the following steps: controlling a moving mechanism carrying an imaging device to move to a detection position, and obtaining a to-be-detected image containing a connecting sheet; inputting the to-be-detected image into a pre-trained target detection model, and obtaining a prediction bounding box of a connection sheet region; performing size expansion based on the prediction bounding box, and generating a region of interest covering the deformation region of the connecting sheet; in the region of interest, extracting a target feature point coordinate of a connecting sheet and a centroid coordinate of a welding spot by using an image processing algorithm; and calculating a distance value between the coordinates of the target feature point and the coordinates of the mass center of the welding spot, and judging the welding quality of the connecting piece based on the distance value. The invention aims to realize high-precision and high-efficiency welding quality detection of the connecting piece under the conditions of complex background interference and uncertain target form.

Description

Technical Field

[0001] This invention relates to the fields of industrial automation and machine vision technology, and in particular to a method, system, medium, and electronic device for inspecting the welding quality of connecting pieces. Background Technology

[0002] In the manufacturing process of lithium-ion battery modules, the welding quality of the connecting pieces and the cell terminals directly affects the electrical performance and mechanical stability of the battery pack. Defects such as poor welding, off-center welding, or over-welding can lead to poor contact or even safety accidents. Currently, quality control on the production line mainly relies on manual visual inspection or traditional machine vision inspection methods.

[0003] However, manual visual inspection is inefficient and makes data traceability difficult. Machine vision solutions based on traditional algorithms are prone to false detection when faced with complex wiring harnesses, bracket textures, and reflective metallic backgrounds inside battery packs. Furthermore, non-rigid deformation often occurs in connecting pieces during processing and welding, leading to a decrease in accuracy or even failure of positioning algorithms based on fixed template matching. In addition, the increasingly stringent requirements of modern production lines for production cycle time make it difficult for existing technologies to simultaneously meet the needs of highly robust target recognition and high-precision geometric measurement within milliseconds.

[0004] Therefore, how to achieve high-precision and high-efficiency welding quality inspection of connecting pieces under complex background interference and uncertain target shape has become an urgent technical problem to be solved. Summary of the Invention

[0005] The main objective of this invention is to provide a method, system, medium, and electronic device for inspecting the welding quality of connecting pieces, aiming to achieve high-precision and high-efficiency inspection of the welding quality of connecting pieces under conditions of complex background interference and uncertain target shape.

[0006] To achieve the above objectives, this invention proposes a method for inspecting the welding quality of connecting pieces, comprising the following steps: The moving mechanism equipped with the imaging device is controlled to move to the detection position and acquire the image to be detected, including the connecting piece; The image to be detected is input into a pre-trained target detection model to obtain the predicted bounding boxes of the connected patch regions. Based on the predicted bounding box, the size is expanded to generate a region of interest that covers the deformed area of ​​the connected sheet. Within the region of interest, image processing algorithms are used to extract the coordinates of the target feature points of the connecting piece and the centroid coordinates of the solder joint; Calculate the distance between the coordinates of the target feature point and the coordinates of the centroid of the weld point, and determine the welding quality of the connecting piece based on the distance value.

[0007] Preferably, the step of controlling the moving mechanism carrying the imaging device to move to the detection position and acquiring the image to be detected, including the connecting piece, specifically includes: Control the end effector of the moving mechanism to move continuously along a preset trajectory; When the moving mechanism passes through the preset shooting point, a hardware trigger signal is sent to the imaging device to trigger the imaging device to perform exposure acquisition and obtain the image to be detected.

[0008] Preferably, before controlling the end of the moving mechanism to move continuously along a preset trajectory, a hand-eye calibration step is further included: The moving mechanism is controlled to carry the imaging device to take pictures of the calibration board in different poses, and the corner coordinates of the calibration board are extracted and solved. The transformation matrix between the coordinate system of the imaging device and the coordinate system of the end tool of the moving mechanism is obtained by using the equation. Based on the transformation matrix, the pixel coordinates in the image acquired by the imaging device are mapped to spatial coordinates in the basic coordinate system of the moving mechanism.

[0009] Preferably, the specific parameters for sending the hardware trigger signal to the imaging device are: Send a TTL hardware trigger signal with a pulse width of 1ms; After receiving the TTL hardware trigger signal, the imaging device performs exposure acquisition with an exposure time of 500μs.

[0010] Preferably, the specific steps of expanding the size based on the predicted bounding box to generate a region of interest covering the deformed region of the connected piece include: Obtain the width of the predicted bounding box. and height ; Calculate extended pixel values ,in equal to the width With the height 10% to 15% of the larger value; Using the center of the predicted bounding box as a reference, the extended pixel value $padding$ is extended outward to generate the region of interest.

[0011] Preferably, the pre-trained target detection model is a YOLOv8 model; the training process of the model includes: The Mosaic data augmentation method was used to process the sample images; Focal Loss is used as the loss function to balance background and foreground samples; The learning rate is adjusted using a cosine annealing strategy.

[0012] Preferably, before extracting the coordinates of the target feature points of the connecting piece and the centroid coordinates of the solder joint using the image processing algorithm, the method further includes: Gaussian filtering is applied to the image within the region of interest to remove noise; The CLAHE algorithm, which uses contrast-limited adaptive histogram equalization, is used to enhance images.

[0013] Preferably, the specific steps for extracting the target feature point coordinates of the connecting piece include: Using a preset standard connection piece template, normalized correlation coefficient matching is performed within the region of interest to obtain the pixel-level optimal matching point; At the pixel-level optimal matching point, sub-pixel interpolation is performed using a quadratic surface fitting method to output the target feature point coordinates with sub-pixel precision.

[0014] Preferably, the specific steps for extracting the centroid coordinates of the solder joint include: Binarization and morphological closing operations are performed on the region of interest to obtain the connected region; Blob analysis is performed on the connected regions to select regions with an area between 100 and 500 pixels and a circularity greater than 0.6 as solder joint regions. Calculate the geometric center of the weld joint region to obtain the centroid coordinates.

[0015] Preferably, the specific steps for calculating the distance between the coordinates of the target feature point and the coordinates of the centroid of the weld point include: The pixel distance between the coordinates of the target feature point and the centroid coordinates of the weld point is calculated according to the Euclidean distance formula. The pixel distance is converted into physical distance using a preset pixel equivalent. The specific steps for determining the welding quality of the connecting piece based on the distance value include: Set a preset first distance threshold, and set a preset second distance threshold; If the physical distance is between the first distance threshold and the second distance threshold, the welding quality of the connecting piece is determined to be qualified; If the physical distance is less than the first distance threshold or greater than the second distance threshold, the welding quality of the connecting piece is determined to be unqualified.

[0016] This application also discloses a welding quality inspection system for connecting pieces, including: A mobile mechanism, with an imaging device mounted at its end, is used to move according to instructions and drive the imaging device to acquire an image to be detected. The processor is communicatively connected to the moving mechanism and the imaging device, and is configured to perform the welding quality inspection method for the connecting piece as described in any one of the above.

[0017] This application also discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the welding quality inspection method for connecting pieces as described in any of the preceding claims.

[0018] This application also discloses an electronic device, comprising: a memory storing a computer program thereon; and a processor for executing the computer program in the memory to implement the welding quality inspection method for connecting pieces as described in any of the preceding claims.

[0019] The above technical solution has the following advantages: This invention achieves rapid coarse localization of the connecting piece region in complex backgrounds by controlling the movement of the imaging device carried by the mobile mechanism and combining it with a pre-trained target detection model. It then generates the region of interest (ROI) based on the size expansion of the predicted bounding box, effectively solving the problem of loss or truncation of key feature regions caused by non-rigid deformation of the connecting piece, ensuring the adaptability and integrity of the subsequent detection range. Furthermore, within the ROI, image processing algorithms are used to extract microscopic features, obtaining high-precision coordinates of target feature points and weld centroids. The welding quality is then quantified by calculating the distance between these two coordinates. This constructs a cascaded detection system of "macroscopic intelligent positioning - microscopic precision measurement," which leverages both the environmental adaptability of deep learning models and the high precision of traditional algorithms, significantly improving the robustness, accuracy, and automation efficiency of connecting piece welding quality detection. Attached Figure Description

[0020] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of a welding quality inspection method for connecting pieces provided in an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the welding quality inspection system for connecting pieces provided in an embodiment of the present invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] Example 1 This embodiment provides a method for inspecting the welding quality of connector pieces, aiming to solve the problem of connector piece inspection on lithium battery PACK production lines in the prior art, especially in situations with complex backgrounds, non-rigid deformation of the target, and extremely high production cycle requirements, where traditional machine vision inspection struggles to simultaneously meet the challenges of high precision and high efficiency. This method is implemented through a cascaded heterogeneous computing framework combining macroscopic inspection and microscopic measurement, as detailed in the specific process below. Figure 1 As shown.

[0024] This detection method first requires establishing a precise spatial reference. The moving mechanism carrying the imaging device is controlled to move to the detection position and acquire the image to be detected, including the connecting piece. Prior to this, the system establishes a mapping relationship between the image pixel coordinate system and the robot's basic coordinate system through a hand-eye calibration step. Specifically, the robot, carrying the imaging device, takes pictures of the calibration board in different poses, and the corner coordinates of the calibration board are extracted and solved... The equation is used to obtain the transformation matrix between the coordinate system of the imaging device and the coordinate system of the robot's end effector. Based on the transformation matrix, the pixel coordinates in the image acquired by the imaging device are mapped to spatial coordinates in the robot's base coordinate system. This process ensures that the pixel distances obtained from subsequent visual measurements can be accurately converted into actual physical distances, providing a mathematical basis for coordinate-based quantitative evaluation.

[0025] In actual online inspection, to match the production cycle of high-speed production lines, this embodiment employs a motion-based imaging mode to acquire images to be inspected. The robot's end effector moves continuously along a pre-planned, smooth path at a speed of 0.5 m / s, covering all the connected pieces to be inspected without collision risk. When the robot passes a pre-defined image capture point, a hardware trigger signal is sent to the imaging device, triggering it to perform exposure and acquire the image to be inspected. Specific parameters are set as follows: a 1 ms TTL hardware trigger signal is sent to the imaging device; upon receiving the TTL hardware trigger signal, the imaging device performs exposure and acquisition with an exposure time of 500 μs. This short-exposure, global shutter shooting method effectively freezes the moving image, avoids motion blur, and allows the robot to complete high-quality image acquisition without stopping during movement, significantly improving inspection efficiency.

[0026] After acquiring the image to be detected, it is input into a pre-trained object detection model to obtain the predicted bounding boxes of the connected patch regions. In this embodiment, the object detection model used is the YOLOv8 model, which has undergone specialized offline training. The training process includes: processing the sample images using the Mosaic data augmentation method to enrich the diversity of the dataset and improve the model's adaptability to different poses and backgrounds; using Focal Loss as the loss function to balance background and foreground samples and solve the problem of imbalanced positive and negative samples; and using a cosine annealing strategy to adjust the learning rate and promote model convergence to the optimal solution. Through this step, the model can quickly and robustly identify the main position of the connected patch in a background containing complex textures such as wire harnesses and supports, and output the center point pixel coordinates. ,width and height The predicted bounding box. This is equivalent to completing the coarse localization at the macro level, defining the range for subsequent fine measurements.

[0027] To address the morphological uncertainty caused by elastic warping or plastic deformation of the connecting piece during welding, this embodiment does not directly rely on the predicted bounding box for measurement. Instead, it expands the dimensions of the predicted bounding box to generate a region of interest (ROI) covering the deformation area of ​​the connecting piece. The specific expansion steps are: obtaining the width of the predicted bounding box... and height ; Calculate extended pixel values ,in equal to width With height 10% to 15% of the larger value.

[0028] For example, take 10%, that is ); Extend the pixel values ​​outward from the center of the predicted bounding box. This generates the region of interest. The specific coordinate calculation formula is as follows:

[0029]

[0030]

[0031]

[0032] This adaptive ROI generation strategy no longer relies on computationally expensive instance segmentation masks. Instead, it uses simple geometric expansion to ensure that even if the connecting piece undergoes a certain degree of non-rigid deformation, its key feature regions can still fall completely within the subsequent precise measurement range, thus effectively avoiding feature loss or truncation caused by target deformation.

[0033] Within the generated region of interest, image processing algorithms are used to extract the coordinates of target feature points on the connector and the centroid coordinates of the solder joints. This process belongs to the microscopic measurement layer, fully leveraging the high precision advantage of traditional image processing algorithms in local areas. Before extracting the coordinates, to further improve image quality, a Gaussian filter with a kernel size of 3x3 is applied to the image within the region of interest to remove noise, and the contrast-limited adaptive histogram equalization (CLAHE) algorithm is used to enhance the image, making the edges of the connector and the texture of the solder joints clearer.

[0034] The specific steps for extracting the target feature point coordinates of the connect piece include: using a preset standard connect piece template, performing normalized correlation coefficient matching within the region of interest to obtain the pixel-level optimal matching point; and at the pixel-level optimal matching point, performing sub-pixel interpolation using a quadratic surface fitting method to output the target feature point coordinates with sub-pixel precision. By employing this coarse-to-fine matching strategy, combined with sub-pixel interpolation technology, a positioning accuracy better than 0.1 pixels was achieved, overcoming the shortcomings of insufficient accuracy in direct pixel matching.

[0035] The specific steps for extracting the centroid coordinates of solder joints include: binarizing the region of interest and performing a morphological closing operation with a kernel size of 3x3 circles to obtain connected regions; performing Blob analysis on the connected regions to select regions with an area between 100 and 500 pixels and a circularity greater than 0.6 as solder joint regions; and calculating the geometric center of the solder joint regions to obtain the centroid coordinates. By setting specific area and roundness thresholds, false solder joint noise such as reflective spots and scratches is effectively filtered out, ensuring that the extracted centroids do indeed correspond to the actual solder joint locations.

[0036] Finally, the distance between the target feature point coordinates and the weld centroid coordinates is calculated, and the welding quality of the connecting piece is determined based on the distance value. Specific steps include: calculating the pixel distance between the target feature point coordinates and the weld centroid coordinates using the Euclidean distance formula. The formula is as follows:

[0037] Convert the pixel distance to the physical distance using a preset pixel equivalent. The specific steps for determining the welding quality of the connecting piece based on the distance value include: setting the first distance threshold to 0.5 mm and the second distance threshold to 1.5 mm; if the physical distance is between the first distance threshold and the second distance threshold, it is determined that the welding quality of the connecting piece is qualified; if the physical distance is less than the first distance threshold or greater than the second distance threshold, it is determined that the welding quality of the connecting piece is unqualified. This determination logic directly reflects the accuracy of the welding position by quantifying the offset degree of the solder joint relative to the characteristic part of the connecting piece. Compared with simply judging whether the solder joint exists, this quantitative evaluation based on relative distance can more sensitively detect hidden defects such as offset welding, virtual welding, and over-welding, thus realizing the closed-loop control of welding quality.

[0038] Embodiment 2 This embodiment provides a connecting piece welding quality detection system, which is the hardware basis for implementing the method described in Embodiment 1. As Figure 2 shown, the system mainly includes a robot 100, an imaging device 200, a processor 300, a light source 400, and an alarm device 500. In this application, the moving mechanism is the robot 100.

[0039] Among them, the robot 100 preferably uses a six-axis industrial robot, such as the Fanuc LR Mate 200iD model, whose repeat positioning accuracy can reach ±0.02 mm, and can meet the trajectory accuracy and stability requirements for high-speed flying shooting. The imaging device 200 is rigidly installed on the end flange of the robot 100. The robot 100 is internally configured with a robot controller 101, which is used to execute a preset motion trajectory program and is responsible for sending an accurate hardware trigger signal to the imaging device 200 according to the motion position.

[0040] The imaging device 200 includes an industrial camera 201 and an optical lens 202. The industrial camera 201 selects a high-resolution global shutter camera, such as a 5-megapixel CMOS camera of the Hikrobot MV-CH050-10UM series, and transmits data to the processor 300 through a GigE interface. The design of the global shutter ensures that when the exposure occurs during the high-speed movement of the robot 100, the image will not produce smear distortion. The optical lens 202 selects a low-distortion telecentric lens, such as a Computar M0820-MPW2 20 mm fixed-focus telecentric lens. The telecentric lens can effectively eliminate perspective errors and ensure the accuracy of the connecting piece size measurement at different depths of field.

[0041] The light source 400 is used to provide a stable lighting environment, preferably using a strip LED white light source and installed at the front end of the imaging device 200 in a low-angle annular lighting mode. This lighting method can effectively highlight the height difference between the edge of the connecting piece and the solder joint and the substrate, forming a clear contour and shadow, which is beneficial for subsequent feature extraction.

[0042] The processor 300 is specifically a high-performance industrial computer, equipped with an Intel Core i7 processor, 32GB of memory, and an NVIDIA GeForce RTX 3060 graphics card. The processor 300 is communicatively connected to the robot 100 and the imaging device 200. Specifically, the imaging device 200 is connected to the processor 300 via a network cable, and the robot controller 101 is connected to the processor 300 via Ethernet for command interaction. The processor 300 is configured to execute the welding quality inspection method for connecting pieces described in Embodiment 1, including loading a pre-trained YOLOv8 model, receiving the image to be inspected, and performing target detection, ROI generation, feature extraction, and quality judgment logic.

[0043] The alarm device 500 is connected to the digital I / O card of the processor 300. When the processor 300 determines that the welding quality of the connecting piece is unqualified, it triggers the alarm device 500 to issue an audible and visual alarm and uploads the defect information to the manufacturing execution system (MES) via network protocol.

[0044] Example 3 This embodiment provides a detailed description of the model training and system calibration processes involved in Embodiment 1 and Embodiment 2, to ensure that those skilled in the art can reproduce the high-precision detection effect.

[0045] Regarding the hand-eye calibration steps, this embodiment uses a high-precision 9x6 checkerboard calibration board with a square size of 3mm. The robot 100, carrying the imaging device 200, is controlled to capture images in at least 15 different poses above the calibration board. These poses include significant translational and rotational changes. After image acquisition, the `findChessboardCorners` function from the OpenCV library is used to extract all corner coordinates at the sub-pixel level. The `calibrateCamera` function is used for camera intrinsic parameter calibration. Finally, the Tsai-Lenz hand-eye calibration algorithm is used to solve for the fixed transformation matrix from the imaging device 200 coordinate system to the robot 100 tool coordinate system, thus establishing the spatial reference.

[0046] Regarding the training of the object detection model, during the system debugging phase, robot 100 was controlled to fly and photograph multiple battery packs of different models, collecting approximately 2000 images of the connected pieces with various poses, lighting conditions, and backgrounds. Boundary boxes were annotated for each connected piece in the images using a labeling tool. On processor 300, the YOLOv8m model was used as pre-trained weights for transfer learning. Training parameters were set as follows: initial learning rate 0.01, batch size 16, and 300 training epochs. Mosaic data augmentation, a cosine annealing learning rate scheduler, and the AdamW optimizer were enabled during training. After training, the model performance was evaluated on the validation set, achieving an average accuracy (mAP@0.5) of 98.5%. The optimal weight file was exported for inference in the online detection phase.

[0047] Furthermore, in the online inspection quality assessment stage, this embodiment sets clear pixel equivalents and physical thresholds. The pre-calibrated pixel equivalent is 0.02 mm / pixel. The set acceptable physical distance interval [D1, D2] is [0.5 mm, 1.5 mm]. When the calculated physical distance D between the feature point and the centroid of the solder joint satisfies 0.5 mm ≤ D ≤ 1.5 mm, it is considered acceptable; if D < 0.5 mm, it is considered a risk of over-soldering; if D > 1.5 mm, it is considered a risk of poor soldering or off-center soldering. This quantitative setting based on specific parameters ensures that the inspection standard is highly consistent with the actual production process specifications.

[0048] This application also discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the welding quality inspection method for connecting pieces as described in any of the preceding claims.

[0049] This application also discloses an electronic device, comprising: a memory storing a computer program thereon; and a processor for executing the computer program in the memory to implement the welding quality inspection method for connecting pieces as described in any of the preceding claims.

[0050] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for inspecting the welding quality of connecting pieces, characterized in that, Includes the following steps: The moving mechanism equipped with the imaging device is controlled to move to the detection position and acquire the image to be detected, including the connecting piece; The image to be detected is input into a pre-trained target detection model to obtain the predicted bounding boxes of the connected patch regions. Based on the predicted bounding box, the size is expanded to generate a region of interest that covers the deformed area of ​​the connected sheet. Within the region of interest, image processing algorithms are used to extract the coordinates of the target feature points of the connecting piece and the centroid coordinates of the solder joint; Calculate the distance between the coordinates of the target feature point and the coordinates of the centroid of the weld point, and determine the welding quality of the connecting piece based on the distance value.

2. The method for inspecting the welding quality of connecting pieces according to claim 1, characterized in that, The step of controlling the moving mechanism carrying the imaging device to move to the detection position and acquiring the image to be detected, including the connecting piece, specifically includes: Control the end effector of the moving mechanism to move continuously along a preset trajectory; When the moving mechanism passes through the preset shooting point, a hardware trigger signal is sent to the imaging device to trigger the imaging device to perform exposure acquisition and obtain the image to be detected.

3. The method for inspecting the welding quality of connecting pieces according to claim 2, characterized in that, Before controlling the end of the moving mechanism to move continuously along a preset trajectory, a hand-eye calibration step is also included: The moving mechanism is controlled to carry the imaging device to take pictures of the calibration board in different poses, and the corner coordinates of the calibration board are extracted and solved. The transformation matrix between the coordinate system of the imaging device and the coordinate system of the end tool of the moving mechanism is obtained by using the equation. Based on the transformation matrix, the pixel coordinates in the image acquired by the imaging device are mapped to spatial coordinates in the basic coordinate system of the moving mechanism.

4. The method for inspecting the welding quality of connecting pieces according to claim 2, characterized in that, The specific parameters for sending the hardware trigger signal to the imaging device are as follows: Send a TTL hardware trigger signal with a pulse width of 1ms; After receiving the TTL hardware trigger signal, the imaging device performs exposure acquisition with an exposure time of 500μs.

5. The method for inspecting the welding quality of connecting pieces according to claim 1, characterized in that, The specific steps for expanding the size of the predicted bounding box to generate a region of interest covering the deformed region of the connected sheet include: Obtain the width of the predicted bounding box. and height ; Calculate extended pixel values ,in equal to the width With the height 10% to 15% of the larger value; Using the center of the predicted bounding box as a reference, the extended pixel value $padding$ is extended outward to generate the region of interest.

6. The method for inspecting the welding quality of connecting pieces according to claim 1, characterized in that, The pre-trained object detection model is a YOLOv8 model; the training process of the model includes: The Mosaic data augmentation method was used to process the sample images; Focal Loss is used as the loss function to balance background and foreground samples; The learning rate is adjusted using a cosine annealing strategy.

7. The method for inspecting the welding quality of connecting pieces according to claim 1, characterized in that, Before extracting the coordinates of the target feature points of the connecting piece and the centroid coordinates of the solder joint using the image processing algorithm, the method further includes: Gaussian filtering is applied to the image within the region of interest to remove noise; The contrast-limited adaptive histogram equalization (CLAHE) algorithm is used to enhance the image.

8. The method for inspecting the welding quality of connecting pieces according to claim 1, characterized in that, The specific steps for extracting the coordinates of the target feature points of the connecting piece include: Using a preset standard connection piece template, normalized correlation coefficient matching is performed within the region of interest to obtain the pixel-level optimal matching point; At the pixel-level optimal matching point, sub-pixel interpolation is performed using a quadratic surface fitting method to output the target feature point coordinates with sub-pixel precision.

9. The method for inspecting the welding quality of connecting pieces according to claim 1, characterized in that, The specific steps for extracting the centroid coordinates of the solder joint include: Binarization and morphological closing operations are performed on the region of interest to obtain the connected region; Blob analysis is performed on the connected regions to select regions with an area between 100 and 500 pixels and a circularity greater than 0.6 as solder joint regions. Calculate the geometric center of the weld joint region to obtain the centroid coordinates.

10. The method for inspecting the welding quality of connecting pieces according to claim 1, characterized in that, The specific steps for calculating the distance between the coordinates of the target feature point and the coordinates of the centroid of the weld point include: The pixel distance between the coordinates of the target feature point and the centroid coordinates of the weld point is calculated according to the Euclidean distance formula. The pixel distance is converted into physical distance using a preset pixel equivalent. The specific steps for determining the welding quality of the connecting piece based on the distance value include: Set a preset first distance threshold, and set a preset second distance threshold; If the physical distance is between the first distance threshold and the second distance threshold, the welding quality of the connecting piece is determined to be qualified; If the physical distance is less than the first distance threshold or greater than the second distance threshold, the welding quality of the connecting piece is determined to be unqualified.

11. A welding quality inspection system for connecting pieces, characterized in that, include: A mobile mechanism, with an imaging device mounted at its end, is used to move according to instructions and drive the imaging device to acquire an image to be detected. The processor, communicatively connected to the moving mechanism and the imaging device, is configured to perform the welding quality inspection method for connecting pieces as described in any one of claims 1 to 10.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the welding quality inspection method for connecting pieces as described in any one of claims 1 to 10.

13. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor is configured to execute the computer program in the memory to implement the welding quality inspection method for connecting pieces as described in any one of claims 1 to 10.

Images