Battery cell pole positioning method and device, milling equipment and storage medium

Abstract

The invention relates to a battery cell pole positioning method and device, milling equipment and a storage medium. The battery cell pole positioning method comprises the following steps: acquiring a color image and a depth image of a target battery acquired by a visual system; mapping the welding ring position detected by the color image, and determining a region of interest in the depth image; performing distance filtering and normal vector consistency filtering on each point cloud in the region of interest; and performing template matching on the filtered region of interest and a standard welding ring point cloud image to obtain a rigid body transformation matrix, and obtaining the three-dimensional coordinate of the circle center of the welding ring of the battery cell pole of the target battery based on the rigid body transformation matrix and the three-dimensional coordinate of the circle center of the welding ring in the standard welding ring point cloud image. According to the cell pole positioning method provided by the invention, point cloud interference under a rough milling working condition can be effectively inhibited, and high-precision and high-robustness extraction of the three-dimensional coordinates of the circle center of the welding ring is realized through an innovative point cloud filtering and fusion registration strategy, so that a robot is accurately guided to finish rough milling operation.

Description

Technical Field

[0001] This invention relates to the field of new energy battery technology, and in particular to a method, apparatus, milling equipment, and storage medium for positioning battery cell terminals. Background Technology

[0002] With the explosive growth of the new energy vehicle industry, the first batch of new energy batteries has entered the stage of large-scale retirement. Automated dismantling, testing, and recycling of retired battery packs is a key link in realizing resource recycling. In the dismantling process, a core step is to use an end mill to rough mill the top cover of the battery pack to expose the cell terminals underneath, thereby picking up the battery pack by the position of the cell terminals and moving it to a designated location for rough milling.

[0003] During rough milling, metal chips and coolant splashes are generated. In addition, the battery cover itself may have irregular surfaces and oil stains. As a result, in the subsequent acquisition of battery pack point cloud images, there are a large number of noise points and outliers at the edge of the target point cloud corresponding to the cell terminal welding ring. These interfering point clouds are mixed with the real welding ring point cloud, which seriously affects the accuracy of subsequent processing.

[0004] Currently, the industry typically uses traditional 3D (three-dimensional) or 2D (two-dimensional) vision methods, such as point cloud segmentation and clustering, for cell electrode positioning. These methods have the following drawbacks: 1. Because interfering point clouds are mixed around the real solder ring point clouds, traditional point cloud processing methods are difficult to effectively separate the interfering point clouds from the real solder ring point clouds, making it difficult to accurately locate the real solder ring position.

[0005] 2. Simple 3D vision methods involve a large amount of data processing and have low real-time performance.

[0006] 3. Simple 2D vision methods cannot directly obtain the accurate three-dimensional coordinate position of the actual weld ring, which is difficult to meet the needs of subsequent precise robot positioning and guidance. Summary of the Invention

[0007] In view of this, it is necessary to provide a method, device, milling equipment and storage medium for positioning battery cell terminals to solve the problems of inaccurate positioning of battery cell terminals, low real-time performance and inability of two-dimensional coordinates to meet the requirements of subsequent processing.

[0008] To address the above problems, this invention provides a method for positioning battery cell terminals, comprising: Acquire color and depth images of the target battery using a vision system; The locations of solder rings detected in the color image are mapped to determine the region of interest in the depth image; For each point cloud in the region of interest, perform distance filtering and normal vector consistency filtering; The filtered region of interest is template-matched with a standard solder ring point cloud image to obtain a rigid body transformation matrix. Based on the rigid body transformation matrix and the three-dimensional coordinates of the solder ring center in the standard solder ring point cloud image, the three-dimensional coordinates of the solder ring center of the cell electrode of the target battery are obtained.

[0009] In one possible implementation, the steps of performing distance filtering and normal vector consistency filtering on each point cloud in the region of interest include: The initial filtered region is obtained by filtering based on the average distance and distance threshold corresponding to each point cloud in the region of interest. The filtered region of interest is obtained by filtering the average angle between the normal vectors of each point cloud in the initial filtering region and the threshold angle between the normal vectors. Wherein, the average distance is the average distance between each point cloud in the region of interest and its K nearest neighbors, and the average angle of the normal vector is the average angle between the normal vector of each point cloud in the initial filtered region and the normal vectors of its K nearest neighbors, where K is a positive integer.

[0010] In one possible implementation, the step of filtering based on the average distance and a distance threshold corresponding to each point cloud in the region of interest to obtain an initial filtered region includes: Calculate the average distance corresponding to each point cloud in the region of interest, and further calculate the first mean and first standard deviation of all average distances; The distance threshold is calculated based on the first mean and the first standard deviation; The point cloud with an average distance greater than the distance threshold in the region of interest is removed to obtain the initial filtered region.

[0011] In one possible implementation, the step of filtering based on the average angle of the normal vectors corresponding to each point cloud in the initial filtering region and a threshold angle of the normal vectors to obtain the filtered region of interest includes: Calculate the average angle of the normal vector corresponding to each point cloud in the initial filtering region, and further calculate the second mean and second standard deviation of the average angle of all normal vectors; The normal vector angle threshold is calculated based on the second mean and the second standard deviation; Point clouds whose average angle of normal vectors in the initial filtered region is greater than the threshold angle of normal vectors are removed to obtain the filtered region of interest.

[0012] In one possible implementation, the normal vector angle threshold is calculated based on the second mean and the second standard deviation, using the following formula: ; in, This represents the threshold angle between the normal vectors. This represents the second mean. This represents the second standard deviation. It is a positive number.

[0013] In one possible implementation, the step of mapping the solder ring positions detected in the color image to determine the region of interest in the depth image includes: The preprocessed color image is used for solder ring identification and positioning to obtain the two-dimensional coordinates of the solder ring region and the center of the solder ring in the color image; Based on the two-dimensional coordinates of the solder ring region and the center of the solder ring, the region of interest is obtained by mapping them onto the depth image.

[0014] In one possible implementation, after obtaining the three-dimensional coordinates of the center of the cell terminal bonding ring of the target battery, the method further includes: The target battery is milled based on the three-dimensional coordinates of the center of the cell terminal weld ring.

[0015] The present invention also provides a battery cell terminal positioning device, comprising: The acquisition module is used to acquire color and depth images of the target battery from the vision system. An interest module is used to map the solder ring positions detected in the color image to determine the region of interest in the depth image; The filtering module is used to perform distance filtering and normal vector consistency filtering on each point cloud in the region of interest; The localization module is used to perform template matching between the filtered region of interest and the standard solder ring point cloud image to obtain a rigid body transformation matrix. Based on the rigid body transformation matrix and the three-dimensional coordinates of the center of the solder ring in the standard solder ring point cloud image, the three-dimensional coordinates of the center of the cell electrode solder ring of the target battery are obtained.

[0016] The present invention also provides a milling machine, including a vision system, a robot, a memory, and a processor, wherein the robot is equipped with an end mill, wherein: The vision system is used to acquire and transmit color and depth images of the target battery; The memory is used to store programs; The processor, coupled to the memory, is used to receive the color image and the depth image, and execute the program stored in the memory to implement the above-mentioned cell electrode positioning method and obtain the three-dimensional coordinates of the center of the cell electrode welding ring of the target battery. The robot is used to control the end mill to move to the position of the three-dimensional coordinates of the center of the circle, and to mill the target battery.

[0017] The present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, enables the above-described cell electrode positioning method.

[0018] The beneficial effects of this invention are as follows: When locating the battery cell terminals, the color image and depth image of the target battery collected by the vision system are first acquired. Based on the color image, the position of the solder ring is coarsely located to obtain the approximate position of the solder ring. Then, the position of the solder ring is mapped to the region of interest in the depth image. By performing a two-dimensional to three-dimensional mapping operation, the point cloud in the entire depth image is avoided, the amount of data processing is reduced, and the real-time performance of the battery cell terminal location method is improved.

[0019] Distance filtering and normal vector consistency filtering are performed on each point cloud in the region of interest. Since the distance between noise points and outliers and other point clouds is large, and the normal vectors of noise points and outliers differ greatly from the normal vectors of the real solder ring point cloud plane, the combination of distance filtering and normal vector consistency filtering can effectively remove noise points and outliers such as metal debris and droplets attached to the edge of the real solder ring. This overcomes the shortcomings of traditional filtering algorithms and improves the anti-interference capability of the battery cell electrode positioning method.

[0020] The filtered region of interest is template-matched with a standard weld ring point cloud image to obtain a rigid body transformation matrix. Based on this rigid body transformation matrix and the 3D coordinates of the weld ring center in the standard weld ring point cloud image, the 3D coordinates of the target battery cell terminal weld ring center are obtained, and these 3D coordinates meet the requirements for subsequent precise robot positioning and guidance. Since the template matching algorithm is not sensitive to point cloud defects or slight deformations, combined with the preceding distance filtering and normal vector consistency filtering, this method can still stably output reliable positioning results even under different levels of interference, thus improving the robustness of this method.

[0021] Furthermore, this invention combines 2D visual coarse positioning with 3D template matching for fine positioning, fully integrating the dual advantages of high resolution of 2D images and depth information provided by 3D point cloud images. This greatly improves the accuracy of the three-dimensional coordinates of the center of the cell electrode welding ring of the target battery, thus enhancing the accuracy of this method.

[0022] In summary, this invention can effectively suppress point cloud interference under rough milling conditions. Through innovative point cloud filtering and fusion registration strategies, it achieves high-precision and robust extraction of the three-dimensional coordinates of the weld ring center, thereby accurately guiding the robot to complete the rough milling operation and improving the efficiency and reliability of automated disassembly of retired battery packs. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an embodiment of a battery cell electrode positioning method provided by the present invention. Figure 2 A point cloud image of a region of interest is provided in one embodiment of the present invention; Figure 3 A filtered point cloud image of the region of interest is provided in one embodiment of the present invention; Figure 4 for Figure 1 A flowchart of a method according to an embodiment of step S103; Figure 5 for Figure 4 A flowchart of a method according to an embodiment of step S401; Figure 6 for Figure 4 A flowchart of a method according to an embodiment of step S402; Figure 7 for Figure 1 A flowchart of a method according to an embodiment of step S102; Figure 8 This is a schematic diagram of an embodiment of a battery cell electrode positioning device provided by the present invention; Figure 9 This is a schematic diagram of an embodiment of a milling machine provided by the present invention. Detailed Implementation

[0024] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0025] A specific embodiment of the present invention, such as Figure 1 As shown, a method for positioning battery cell terminals is disclosed, including: The embodiments of the present invention can be applied to various scenarios that require positioning of the cell terminals. Common application scenarios include disassembling retired battery packs and welding and assembling individual new energy batteries.

[0026] For example, in the process of disassembling a retired battery pack, the hardware involved includes a vision system, a robot, and a server. The vision system can be fixed at the end of the robot or above the processing scene to ensure that the top cover of the target battery to be disassembled is within the camera's field of view, so that the vision system can clearly capture color and depth images of the battery pack top cover. The end of the robot is equipped with a milling cutter for rough milling the terminal weld rings of the target battery. Communication between any two of the vision system, the robot, and the server can be performed as needed, either via wired or wireless communication.

[0027] Before using this cell electrode positioning method, certain preparatory work needs to be done offline, such as calibrating the camera's intrinsic and extrinsic parameters, establishing the transformation relationship between image pixel coordinates, camera coordinate system and robot coordinate system; it is also necessary to collect point cloud images of standard weld ring samples with no edge interference, which are the standard weld ring point cloud images.

[0028] When using this cell electrode positioning method, the robot first moves to a preset position. After the robot moves to the preset position, it triggers the vision system to take a picture and collect a color image and a depth image of the target battery. After the vision system collects the color image and the depth image, it sends them to the server. After receiving the color image and the depth image, the server executes this cell electrode positioning method to obtain the three-dimensional coordinates of the center of the cell electrode welding ring of the target battery and sends them to the robot.

[0029] After receiving the three-dimensional coordinates of the center of the cell terminal weld ring of the target battery, the robot transforms the three-dimensional coordinates from the camera coordinate system to the robot coordinate system, and then controls the end mill to move to the corresponding position to mill the cell terminal weld ring of the target battery.

[0030] It should be noted that the target battery can be one or multiple. When there are multiple target batteries, the robot controls the end mill to mill each weld ring in sequence.

[0031] The vision system can be a three-dimensional sensor that integrates color images and depth information. For example, the vision system can be a Keinect camera or a ToF (Time-of-Flight) camera, or it can be a composite system consisting of a 2D industrial camera and a 3D laser profilometer, or a 2D industrial camera and a 3D structured light camera.

[0032] The robot can be a six-axis industrial robot, such as FANUC or KUKA, with an electric spindle and end mill as the end effectors.

[0033] In this embodiment, the server is the execution entity. The server can be a terminal device with data processing function, such as a server, smartphone, tablet computer, portable computer, or desktop computer. The server can be a rack server, blade server, tower server, or cabinet server. The specific type can be determined according to the actual situation, and this embodiment does not make any specific limitation.

[0034] S101, acquire color and depth images of the target battery from the vision system; In this embodiment of the invention, the vision system starts working after being triggered, and captures images of the working area where the target battery is located, obtaining color and depth images. The triggering condition for the vision system can be determined according to actual conditions. For example, when the robot moves to a preset position, the robot or the server sends a trigger command to the vision system, and the vision system begins capturing images upon receiving the trigger command.

[0035] The vision system sends the acquired color and depth images to the server. For example, the vision system and the server can be connected via a wired connection, such as a wired cable, to send the color and depth images to the server, which then receives them. Alternatively, the vision system can communicate with the server wirelessly, such as via Bluetooth or Wi-Fi, to send the color and depth images to the server, which then receives them.

[0036] S102, Map the solder ring positions detected by the color image to determine the region of interest in the depth image; After receiving the color image and depth image, the server first performs two-dimensional detection and recognition on the solder ring position in the color image to preliminarily identify the approximate position of the solder ring in the image. Then, it maps the solder ring position to the depth image and extracts a three-dimensional point cloud region of interest (ROI) containing the solder ring and a small amount of background area around it. This mapping process avoids processing the point cloud data of the entire depth image. Subsequently, only the point cloud data of the ROI needs to be focused on, which greatly reduces the amount of data processing and improves the real-time performance of the cell electrode positioning method.

[0037] like Figure 2 As shown, different colors in the region of interest represent different distances from the visual system. The white part at the very center indicates the absence of point clouds, while blue, yellow, and red represent distances from the visual system from far to near.

[0038] S103, perform distance filtering and normal vector consistency filtering on each point cloud in the region of interest; Due to metal debris and coolant splashes generated during rough milling, as well as potential irregularities and oil contamination on the battery cover itself, the edges of the actual solder rings in the region of interest contain numerous noise points and outliers. Analysis revealed that traditional methods based on global distribution characteristics, such as statistical filtering, while removing noise, are prone to inadvertently damaging the effective point cloud at the actual solder ring edges, leading to feature loss. Clustering methods, on the other hand, are sensitive to initial parameters and struggle to effectively separate noise points when they are close to the actual solder rings. Therefore, traditional point cloud preprocessing methods such as voxel mesh downsampling, statistical filtering, and radius filtering are ineffective in handling such scenarios.

[0039] In response, embodiments of the present invention use distance filtering and normal vector consistency filtering to separate real solder rings from noise points and outliers.

[0040] Since noise points are usually isolated, they are typically far from other point clouds. This embodiment of the invention uses distance filtering to remove noise points that are obviously isolated and far from the real solder ring point cloud. Furthermore, since the surface of an ideal solder ring should be a continuous and smooth plane with a uniformly changing normal vector, and the normal vectors of noise points and outliers differ greatly from the normal vector of this plane, normal vector consistency filtering can remove noise points and outliers that are confused around the real solder ring point cloud.

[0041] In this embodiment of the invention, distance filtering and normal vector consistency filtering are sequentially performed on each point cloud in the region of interest. Since the distance between noise points and outliers and other point clouds is large, and the normal vectors of noise points and outliers differ greatly from the normal vectors of the real solder ring point cloud plane, the combination of distance filtering and normal vector consistency filtering can effectively remove noise points and outliers such as metal debris and droplets attached to the edge of the real solder ring. This overcomes the shortcomings of traditional filtering algorithms and improves the anti-interference capability of the cell electrode positioning method.

[0042] like Figure 3 As shown, noise points and outliers were removed from the filtered region of interest.

[0043] To verify this, the method can effectively remove noise points and outliers by comparing the region of interest and the filtered region of interest in Table 1 in the embodiments of the present invention.

[0044] Table 1

[0045] S104, perform template matching between the filtered region of interest and the standard solder ring point cloud image to obtain a rigid body transformation matrix. Based on the rigid body transformation matrix and the three-dimensional coordinates of the center of the solder ring in the standard solder ring point cloud image, obtain the three-dimensional coordinates of the center of the cell electrode solder ring of the target battery.

[0046] Finally, template matching is performed between the filtered region of interest and the standard weld ring point cloud image to obtain the rigid body transformation matrix. For example, the Iterative Closest Points (ICP) algorithm or an improved algorithm such as feature-based ICP is used for fine-grained template matching to calculate the optimal spatial transformation matrix from the standard weld ring point cloud image to the filtered region of interest. This optimal spatial transformation matrix is ​​the rigid body transformation matrix, which includes a rotation matrix and a translation vector.

[0047] The essence of template matching using the ICP algorithm is to find an optimal rigid body transformation that minimizes the distance between the standard weld ring point cloud image and the filtered region of interest. This problem can be formulated as: ; in, It is the first in the standard weld ring point cloud image A point cloud, It is the first in the filtered region of interest. The nearest neighbor point corresponds to a point cloud, where the nearest neighbor point is the point cloud closest to all its neighbors. Represents the rotation matrix. Represents the translation vector. This represents a standard weld ring point cloud image. This represents the filtered region of interest. It is a positive integer.

[0048] The optimal rotation matrix and translation vector can be efficiently obtained using methods such as Singular Value Decomposition (SVD). Then, the 3D coordinates of the weld ring center in the standard weld ring point cloud image are transformed to the target coordinate system using the rotation matrix and translation vector. This is: ; in, This represents the final, precise three-dimensional coordinates of the center of the target battery cell terminal weld ring, obtained from the solution. This represents the three-dimensional coordinates of the center of the weld ring in a standard weld ring point cloud image.

[0049] During implementation, the three-dimensional coordinates of the center of the weld ring, which are predefined on the standard weld ring point cloud image in the offline state, are obtained by transforming the rigid body transformation matrix obtained in the above steps. This yields the three-dimensional coordinates of the center of the weld ring of the target battery cell electrode.

[0050] In this embodiment of the invention, a template matching method is used. The template matching algorithm is not sensitive to partial missing points or slight deformations in the point cloud. Combined with the previous distance filtering and normal vector consistency filtering, this method can still output reliable positioning results stably even under different degrees of interference, thus improving the robustness of the method.

[0051] Compared with the prior art, the beneficial effects of this embodiment are as follows: When locating the battery cell terminals, the color image and depth image of the target battery collected by the vision system are first acquired. Based on the color image, the position of the solder ring is coarsely located to obtain the approximate position of the solder ring. Then, the position of the solder ring is mapped to the region of interest in the depth image. By performing a two-dimensional to three-dimensional mapping operation, the point cloud in the entire depth image is avoided, the amount of data processing is reduced, and the real-time performance of the battery cell terminal location method is improved.

[0052] Distance filtering and normal vector consistency filtering are performed on each point cloud in the region of interest. Since the distance between noise points and outliers and other point clouds is large, and the normal vectors of noise points and outliers differ greatly from the normal vectors of the real solder ring point cloud plane, the combination of distance filtering and normal vector consistency filtering can effectively remove noise points and outliers such as metal debris and droplets attached to the edge of the real solder ring. This overcomes the shortcomings of traditional filtering algorithms and improves the anti-interference capability of the battery cell electrode positioning method.

[0053] The filtered region of interest is template-matched with a standard weld ring point cloud image to obtain a rigid body transformation matrix. Based on this rigid body transformation matrix and the 3D coordinates of the weld ring center in the standard weld ring point cloud image, the 3D coordinates of the target battery cell terminal weld ring center are obtained, and these 3D coordinates meet the requirements for subsequent precise robot positioning and guidance. Since the template matching algorithm is not sensitive to point cloud defects or slight deformations, combined with the preceding distance filtering and normal vector consistency filtering, this method can still stably output reliable positioning results even under different levels of interference, thus improving the robustness of this method.

[0054] Furthermore, in this embodiment of the invention, the method of coarse positioning using 2D vision combined with fine positioning using 3D template matching fully integrates the dual advantages of high resolution of 2D images and depth information provided by 3D point cloud images, greatly improving the accuracy of the three-dimensional coordinates of the center of the cell electrode welding ring of the target battery and enhancing the accuracy of this method.

[0055] In summary, the embodiments of the present invention can effectively suppress point cloud interference under rough milling conditions. Through innovative point cloud filtering and fusion registration strategies, high-precision and robust extraction of the three-dimensional coordinates of the weld ring center is achieved, thereby accurately guiding the robot to complete the rough milling operation and improving the efficiency and reliability of automated disassembly of retired battery packs.

[0056] In some possible embodiments of the present invention, such as Figure 4 As shown, in step S103, the step of performing distance filtering and normal vector consistency filtering on each point cloud in the region of interest includes: S401, filter according to the average distance and distance threshold corresponding to each point cloud in the region of interest to obtain an initial filtered region, wherein the average distance is the average distance between each point cloud in the region of interest and its K nearest neighbors; Considering that noise points are usually isolated and therefore far from other point clouds, the average distance for each point cloud in the region of interest is calculated. For example, the average distance is the average of the distances between the point cloud and its K nearest neighbors. For each of the K nearest neighbors, the Euclidean distance between the point cloud and each nearest neighbor is calculated, and the average of all Euclidean distances is taken as the average distance for the point cloud.

[0057] It should be noted that the K nearest neighbors refer to the K nearest points in the point cloud, where K is a positive integer.

[0058] The average distance corresponding to each point cloud is compared with the distance threshold. Based on the comparison results, the point clouds are filtered, and those that do not meet the requirements are removed to obtain the initial filtering region.

[0059] S402, filtering is performed based on the average angle between the normal vectors of each point cloud in the initial filtering region and the normal vector angle threshold to obtain the filtered region of interest. The average angle between the normal vectors is the average value of the angles between the normal vector of each point cloud in the initial filtering region and the normal vectors of its K nearest neighbors.

[0060] Considering that the surface of an ideal weld ring should be a continuous and smooth plane with uniform changes in its normal vector direction, while the normal vectors of noise points and outliers differ greatly from the normal vector of the main plane, noise points and outliers whose normal vectors are inconsistent with the main plane can be eliminated by calculating the angle or difference between the normal vectors of each point cloud and its K neighboring points.

[0061] For example, the normal vector of each point cloud can be determined by fitting a local plane formed by the K nearest neighbors of the point cloud. The normal vector of this local plane can be regarded as the normal vector of the point cloud. The normal vectors of the K nearest neighbors are determined in the same way. For each of these K nearest neighbors' normal vectors, the angle between the normal vector of the point cloud and the normal vector of each neighbor is calculated. This angle can be obtained by calculating the cosine of the two normal vectors through the dot product. The average of all the angles is taken as the average angle of the normal vector corresponding to the point cloud.

[0062] It should be noted that the K-nearest neighbor refers to the point cloud surrounding the point cloud.

[0063] The average angle of the normal vectors corresponding to each point cloud is compared with the normal vector angle threshold. Based on the comparison results, the point clouds are filtered, and those that do not meet the requirements are removed to obtain the filtered region of interest.

[0064] In some possible embodiments of the present invention, such as Figure 5 As shown, in step S401, the step of filtering based on the average distance and distance threshold corresponding to each point cloud in the region of interest to obtain the initial filtered region includes: S501, calculate the average distance corresponding to each point cloud in the region of interest, and further calculate the first mean and first standard deviation of all average distances; In practical implementation, the pcl::KdTree function in the Point Cloud Library (PCL) can be used to perform nearest neighbor search for each point cloud. By inputting the three-dimensional coordinates of each point cloud into the pcl::KdTree function, the average distance between each point cloud and its K nearest neighbors can be obtained, which is the average distance corresponding to each point cloud.

[0065] The calculation formula is as follows: ; in, Indicates the first region of interest The average distance corresponding to each point cloud Indicates the first region of interest A point cloud, Indicates the first region of interest Point cloud.

[0066] S502, calculate the distance threshold based on the first mean and the first standard deviation; For example, the formula for calculating the distance threshold in an embodiment of the present invention can be as follows: ; in, Indicates the distance threshold. This represents the first mean. Indicates the first standard deviation. This is a coefficient, typically ranging from 1.0 to 3.0, and can be adjusted according to the actual situation.

[0067] S503, remove point clouds whose average distance in the region of interest is greater than the distance threshold to obtain the initial filtered region.

[0068] For each point cloud in the region of interest, if the average distance of the point cloud is greater than the distance threshold, the point cloud can be determined as a noise point and removed; otherwise, the point cloud is retained, thus obtaining the initial filtering region.

[0069] As an optional example, such as Figure 6As shown, step S402, which involves filtering based on the average angle of the normal vectors corresponding to each point cloud in the initial filtering region and a threshold angle of the normal vectors to obtain the filtered region of interest, includes the following steps: S601, calculate the average angle of the normal vector corresponding to each point cloud in the initial filtering region, and further calculate the second mean and second standard deviation of the average angle of all normal vectors. For example, for each point cloud in the initial filtering region, a local plane formed by the K nearest neighbors of the point cloud is fitted, and the normal vector of this local plane is used as the normal vector of the point cloud. The angle between the normal vector of the point cloud and the normal vector of each of the K nearest neighbors is calculated. This angle can be obtained by calculating the cosine of the angle using the dot product method. The average of all angles is the average angle of the normal vector corresponding to the point cloud. Further, the second mean and second standard deviation of the average angles of all normal vectors are calculated.

[0070] S602, calculate the normal vector angle threshold based on the second mean and the second standard deviation; For example, the normal vector angle threshold is calculated based on the second mean and the second standard deviation, using the following formula: ; in, This represents the threshold angle between the normal vectors. This represents the second mean. This represents the second standard deviation. It is a positive number, usually ranging from 1.0 to 3.0, and can be adjusted according to the actual situation.

[0071] S603, remove point clouds in the initial filtered region where the average angle of the normal vectors is greater than the normal vector angle threshold to obtain the filtered region of interest.

[0072] The average angle of the normal vectors corresponding to each point cloud in the initial filtering region is compared with the normal vector angle threshold. If the average angle of the normal vectors of the point cloud is greater than the normal vector angle threshold, the point cloud can be determined as a noise point and removed. Otherwise, the point cloud is retained, and the filtered region of interest is finally obtained.

[0073] In some possible embodiments of the present invention, in step S102, such as Figure 7 As shown, the step of mapping the solder ring positions detected in the color image to determine the region of interest in the depth image includes: S701, The preprocessed color image is used for solder ring identification and positioning to obtain the two-dimensional coordinates of the solder ring area and the center of the solder ring in the color image; The color image is preprocessed, such as by denoising and enhancing, to obtain a preprocessed color image. Then, the solder rings in the color image are detected to obtain the two-dimensional coordinates of the solder ring region and the center of the solder ring in the color image. The solder ring region represents the approximate area of ​​the solder ring in the two-dimensional color image.

[0074] For example, a deep learning-based image segmentation model can be used for weld ring detection, such as the U-Net network, which is named for its unique U-shaped structure. This structure consists of two parts: an encoder and a decoder, connected by a skip connection. Alternatively, image processing algorithms such as edge detection and Hough transform can be used for weld ring detection.

[0075] S702, based on the two-dimensional coordinates of the solder ring region and the center of the solder ring, the region of interest is obtained by mapping them onto the depth image.

[0076] The two-dimensional coordinates of the weld ring region and its center are mapped onto the depth image, and a three-dimensional point cloud region of interest containing the target weld ring and a small amount of surrounding background area is extracted.

[0077] In some possible embodiments of the present invention, after obtaining the three-dimensional coordinates of the center of the cell terminal bonding ring of the target battery, the method further includes: The target battery is milled based on the three-dimensional coordinates of the center of the cell terminal weld ring.

[0078] The method for milling the target battery based on the three-dimensional coordinates of the center of the target battery terminal weld ring is the same as in the above embodiments. For details, please refer to the above implementation methods. The embodiments of the present invention will not be described in detail here.

[0079] A specific embodiment of the present invention, such as Figure 8 As shown, a battery cell electrode positioning device 800 is disclosed, comprising: The acquisition module 801 is used to acquire color and depth images of the target battery acquired by the vision system; Interest module 802 is used to map the position of the solder ring detected by the color image to determine the region of interest in the depth image; Filtering module 803 is used to perform distance filtering and normal vector consistency filtering on each point cloud in the region of interest; The positioning module 804 is used to perform template matching between the filtered region of interest and the standard solder ring point cloud image to obtain a rigid body transformation matrix. Based on the rigid body transformation matrix and the three-dimensional coordinates of the center of the solder ring in the standard solder ring point cloud image, the three-dimensional coordinates of the center of the cell electrode solder ring of the target battery are obtained.

[0080] This embodiment is a device embodiment corresponding to the above method. Its specific implementation process is the same as that of the above method embodiment. For details, please refer to the above method embodiment. This device embodiment will not repeat the details.

[0081] Each module in the aforementioned cell electrode positioning device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0082] In one embodiment, such as Figure 9 As shown, the present invention also provides a milling machine 900. The milling machine 900 includes a vision system 901, a robot 902, a memory 903, and a processor 904, wherein an end mill is mounted on the robot. Figure 9 Only some components of the milling machine 900 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0083] The vision system 901, robot 902, and end mill 905 have been fully described in the above method embodiments. For details, please refer to the above method embodiments. This embodiment will not repeat them here.

[0084] In some embodiments, memory 903 may be an internal storage unit of the milling machine 900, such as a hard disk or memory of the milling machine 900. In other embodiments, memory 903 may also be an external storage device of the milling machine 900, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the milling machine 900.

[0085] Furthermore, the memory 903 may include both internal storage units of the milling machine 900 and external storage devices. The memory 903 is used to install the application software and various data of the milling machine 900.

[0086] In some embodiments, processor 904 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 904 may be local or remote. In some embodiments, processor 904 may be implemented on a cloud platform. In some embodiments, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, or any combination thereof.

[0087] Furthermore, when the processor 904 executes the cell terminal positioning program in the memory 903, the following steps can be implemented: Acquire color and depth images of the target battery using a vision system; The locations of solder rings detected in the color image are mapped to determine the region of interest in the depth image; For each point cloud in the region of interest, perform distance filtering and normal vector consistency filtering; The filtered region of interest is template-matched with a standard solder ring point cloud image to obtain a rigid body transformation matrix. Based on the rigid body transformation matrix and the three-dimensional coordinates of the solder ring center in the standard solder ring point cloud image, the three-dimensional coordinates of the solder ring center of the cell electrode of the target battery are obtained.

[0088] It should be understood that when the processor 904 executes the program for the cell terminal positioning method in the memory 903, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0089] Accordingly, embodiments of the present invention also provide a computer storage medium storing a computer program, which, when executed by a processor, implements the steps of the cell electrode positioning method described above. Alternatively, when executed by a processor, the computer program implements the functions of each module / unit in the aforementioned embodiment of the cell electrode positioning device.

[0090] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0091] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0092] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for positioning battery cell terminals, characterized in that, include: Acquire color and depth images of the target battery using a vision system; The locations of solder rings detected in the color image are mapped to determine the region of interest in the depth image; For each point cloud in the region of interest, perform distance filtering and normal vector consistency filtering; The filtered region of interest is template-matched with a standard solder ring point cloud image to obtain a rigid body transformation matrix. Based on the rigid body transformation matrix and the three-dimensional coordinates of the solder ring center in the standard solder ring point cloud image, the three-dimensional coordinates of the solder ring center of the cell electrode of the target battery are obtained.

2. The cell electrode positioning method according to claim 1, characterized in that, The steps of performing distance filtering and normal vector consistency filtering on each point cloud in the region of interest include: The initial filtered region is obtained by filtering based on the average distance and distance threshold corresponding to each point cloud in the region of interest. The filtered region of interest is obtained by filtering the average angle between the normal vectors of each point cloud in the initial filtering region and the threshold angle between the normal vectors. Wherein, the average distance is the average distance between each point cloud in the region of interest and its K nearest neighbors, and the average angle of the normal vector is the average angle between the normal vector of each point cloud in the initial filtered region and the normal vectors of its K nearest neighbors, where K is a positive integer.

3. The cell electrode positioning method according to claim 2, characterized in that, The step of filtering based on the average distance and distance threshold corresponding to each point cloud in the region of interest to obtain the initial filtered region includes: Calculate the average distance corresponding to each point cloud in the region of interest, and further calculate the first mean and first standard deviation of all average distances; The distance threshold is calculated based on the first mean and the first standard deviation; The point cloud with an average distance greater than the distance threshold in the region of interest is removed to obtain the initial filtered region.

4. The cell electrode positioning method according to claim 2, characterized in that, The step of filtering based on the average angle of the normal vectors corresponding to each point cloud in the initial filtered region and a threshold angle of the normal vectors to obtain the filtered region of interest includes: Calculate the average angle of the normal vector corresponding to each point cloud in the initial filtering region, and further calculate the second mean and second standard deviation of the average angle of all normal vectors; The normal vector angle threshold is calculated based on the second mean and the second standard deviation; Point clouds whose average angle of normal vectors in the initial filtered region is greater than the threshold angle of normal vectors are removed to obtain the filtered region of interest.

5. The cell electrode positioning method according to claim 4, characterized in that, The normal vector angle threshold is calculated based on the second mean and the second standard deviation, using the following formula: ； in, This represents the threshold angle between the normal vectors. This represents the second mean. This represents the second standard deviation. It is a positive number.

6. The cell electrode positioning method according to claim 1, characterized in that, The step of mapping the solder ring positions detected in the color image to determine the region of interest in the depth image includes: The preprocessed color image is used for solder ring identification and positioning to obtain the two-dimensional coordinates of the solder ring region and the center of the solder ring in the color image; Based on the two-dimensional coordinates of the solder ring region and the center of the solder ring, the region of interest is obtained by mapping them onto the depth image.

7. The cell electrode positioning method according to claim 1, characterized in that, After obtaining the three-dimensional coordinates of the center of the cell terminal bonding ring of the target battery, the method further includes: The target battery is milled based on the three-dimensional coordinates of the center of the cell terminal weld ring.

8. A battery cell electrode positioning device, characterized in that, include: The acquisition module is used to acquire color and depth images of the target battery from the vision system. An interest module is used to map the solder ring positions detected in the color image to determine the region of interest in the depth image; The filtering module is used to perform distance filtering and normal vector consistency filtering on each point cloud in the region of interest; The localization module is used to perform template matching between the filtered region of interest and the standard solder ring point cloud image to obtain a rigid body transformation matrix. Based on the rigid body transformation matrix and the three-dimensional coordinates of the center of the solder ring in the standard solder ring point cloud image, the three-dimensional coordinates of the center of the cell electrode solder ring of the target battery are obtained.

9. A milling machine, characterized in that, Includes a vision system, a robot, a memory, and a processor, wherein the robot is equipped with an end mill, wherein: The vision system is used to acquire and transmit color and depth images of the target battery; The memory is used to store programs; The processor, coupled to the memory, is used to receive the color image and the depth image, and execute the program stored in the memory to implement the cell electrode positioning method according to any one of claims 1 to 7, and to obtain the three-dimensional coordinates of the center of the cell electrode welding ring of the target battery. The robot is used to control the end mill to move to the position of the three-dimensional coordinates of the center of the circle, and to mill the target battery.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, enable the cell terminal positioning method as described in any one of claims 1 to 7.

Images