Integrated battery production system based on visual inspection

Abstract

The invention discloses a battery integrated production system based on visual inspection, and belongs to the technical field of battery processing, and the system comprises a sensing layer which is deployed on a battery production line and is used for collecting production data of a battery, and the production data comprises image data, equipment state data and process parameter data; the control layer is used for performing defect and size detection on the battery according to the image data to obtain a detection result, and generating a control instruction according to the equipment state data, the process parameter data and the detection result; the execution layer is used for responding to the control instruction generated by the control layer and executing battery production operation, and the production operation comprises rolling operation, transmission operation and sorting operation; and the management layer is used for carrying out storage, traceability management and visual display on the battery production data collected by the sensing layer and the detection result detected by the control layer. According to the invention, the production efficiency, the quality stability and the traceability are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of battery processing, and particularly relates to a battery integrated production system based on visual detection. BACKGROUND

[0002] With the rapid development of new energy vehicles, energy storage equipment and other fields, the market demand for batteries is growing explosively, and higher requirements are put forward for the quality, efficiency and safety of battery production. As a key link in battery production, the precision and stability of the battery rolling process directly affect the performance and service life of the battery. The traditional production line relies on manual quality detection and data recording, which has problems such as low detection efficiency, large subjective error and difficult quality traceability, and is difficult to meet the production demand of large scale and high precision. SUMMARY

[0003] The purpose of the present application is to provide a battery integrated production system based on visual detection to solve the problems of low detection efficiency, large subjective error and difficult quality traceability in the prior art.

[0004] In order to achieve the above-mentioned purpose, the application adopts the following technical scheme:

[0005] The application provides a battery integrated production system based on visual detection, which comprises:

[0006] A perception layer is arranged on the battery production line and used to collect production data of the battery, wherein the production data comprises image data, equipment state data and process parameter data;

[0007] A control layer is used to detect defects and sizes of the battery according to the image data, obtain a detection result, and generate a control instruction according to the equipment state data, the process parameter data and the detection result;

[0008] An execution layer is used to execute production operations of the battery in response to the control instruction generated by the control layer, wherein the production operations comprise rolling operation, transmission operation and sorting operation;

[0009] A management layer is used to store, trace and visually display the production data of the battery collected by the perception layer and the detection result detected by the control layer.

[0010] Preferably, the execution layer comprises a rolling device, a transmission device and a sorting device, and the control instruction comprises a rolling instruction, a transmission instruction and a sorting instruction;

[0011] The rolling device is used to execute the rolling operation of the battery in response to the rolling instruction;

[0012] The transmission device is used to execute the transmission operation of the battery in response to the transmission instruction;

[0013] The sorting device is used for performing a sorting operation on the battery in response to a sorting instruction.

[0014] Preferably, the step of traceability management of the management layer on the production data of the battery collected by the perception layer and the detection result detected by the control layer comprises:

[0015] Obtaining a unique identifier pre-assigned to each battery;

[0016] Based on the production timeline, the unique identifier of each battery is associated with the corresponding equipment state data, process parameter data and detection result to obtain association data;

[0017] Obtaining the identity information of the operator, and binding the identity information with the association data to obtain traceability data.

[0018] Preferably, the step of defect detection of the control layer on the battery comprises:

[0019] Pretreating the image data to obtain a target image;

[0020] Feature extraction is performed on the target image to obtain feature information;

[0021] The feature information is input into a pre-constructed defect recognition model for defect prediction to obtain the defect type of the battery, and the defect type of the battery is taken as the detection result.

[0022] Preferably, the image data is two-view images collected by a binocular camera, and the step of size detection of the control layer on the battery comprises:

[0023] Calibrating the binocular camera based on a preset algorithm to obtain a calibration result;

[0024] Using the calibration result to perform stereo correction on the two-view images to obtain corrected images of the two views;

[0025] Stereo matching is performed on the corrected images of the two views to obtain a disparity map;

[0026] Obtaining the focal length and baseline length of the binocular camera, and determining a depth map according to the focal length and baseline length of the binocular camera and the disparity map, the depth map being used to represent the three-dimensional coordinates of each pixel point in the two-dimensional image;

[0027] According to the depth map, the size information of the battery is determined, and the size information of the battery at least includes the length, width and thickness of the battery, and the size information of the battery is taken as the detection result.

[0028] Preferably, the preset algorithm is Zhang's calibration method, and the step of calibrating the binocular camera based on the preset algorithm comprises:

[0029] The two cameras of the binocular camera are calibrated respectively by Zhang's calibration method to obtain the internal parameters and external parameters of each camera, the internal parameters including focal length, principal point coordinates and distortion parameters, and the external parameters including rotation matrix and translation vector;

[0030] Coordinate systems of the two cameras are constructed and converted to the same world coordinate system;

[0031] In the world coordinate system, the relative position and attitude between the two cameras are determined according to the rotation matrix and translation vector between the two cameras.

[0032] Preferably, the step of performing stereo matching on the corrected images of the two views comprises:

[0033] Any pixel point in the corrected images of the two views is selected as a target pixel point, and a pixel window of the target pixel point is constructed, the center pixel of the pixel window being the target pixel point;

[0034] A reference pixel point of the pixel window is determined based on a preset selection rule;

[0035] A matching cost of the target pixel point is determined based on the reference pixel point and each pixel point in the pixel window;

[0036] The matching cost of the target pixel point is corrected based on a preset correction rule to obtain a corrected matching cost of the target pixel point; each pixel point in the corrected images of the two views is traversed to obtain a corrected matching cost corresponding to each pixel point;

[0037] The corrected matching costs corresponding to each pixel point are aggregated to obtain a disparity map.

[0038] Preferably, the preset selection rule is: calculating an average pixel value in the pixel window, calculating an absolute pixel difference between the average pixel value and the pixel value of the target pixel point; determining whether the absolute pixel difference is less than a preset value, if yes, taking the pixel value of the target pixel point as the pixel value of the reference pixel point; if no, taking the average pixel value as the pixel value of the reference pixel point.

[0039] Preferably, the preset correction rule is:

[0040] A change rate between each pixel point in each pixel window and the reference pixel point is calculated;

[0041] The change rates corresponding to each pixel point in the two pixel windows of the same pixel window corresponding to the pixel points at the same position in the two corrected images are subtracted to obtain a change difference value; all change difference values are accumulated and summed to obtain a cost correction value, and the matching cost corresponding to the pixel window is corrected by using the cost correction value to obtain the corrected matching cost of the target pixel point in the pixel window.

[0042] The beneficial effects of the present application are:

[0043] 1、The battery integrated production system comprises four layers of framework, namely sensing, control, execution and management, the sensing layer collects image data, equipment state data and process parameter data, the control layer receives the data transmitted by the sensing layer and controls the execution layer such as the rolling equipment and the transmission equipment in real time; the execution layer responds to the control instruction to execute the production operation of the battery, and the management layer stores, traces and visually displays the data generated in the production process; the system realizes real-time detection, data binding and intelligent tracing of the production process, and improves the production efficiency, quality stability and traceability;

[0044] 2、The control layer also detects defects and sizes of the battery according to the image data, so that the surface defects and size deviations of the battery can be found in time, and the defective products can be removed by the automatic sorting equipment, so that the defective products are prevented from flowing into the next process. BRIEF DESCRIPTION OF DRAWINGS

[0045] The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the embodiments of the present application together with the following specific embodiments, but do not constitute a limitation on the embodiments of the present application. In the drawings:

[0046] Figure 1 is a block diagram of a battery integrated production system based on visual detection provided by an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the present application will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the drawings structure is only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor. It should be noted that the description of these embodiment modes is used to help understand the present application, but does not constitute a limitation on the present application.

[0048] Figure 1 is a block diagram of a battery integrated production system based on visual detection provided by an embodiment of the present application. As Figure 1As shown, this embodiment provides a vision-based integrated battery production system, which consists of a perception layer, a control layer, an execution layer, and a management layer.

[0049] The sensing layer is deployed on the battery production line to collect battery production data, which includes image data, equipment status data, and process parameter data.

[0050] In this embodiment, the perception layer consists of machine vision cameras and various sensors (such as pressure sensors, temperature sensors, displacement sensors, etc.), responsible for collecting image data, equipment status data, and process parameter data during the production process. The machine vision cameras are deployed at key nodes such as the rolling station and inspection station to acquire real-time image data of the battery's appearance and dimensions, while the sensors monitor real-time process parameter data such as rolling pressure, temperature, and speed. The equipment status data in this embodiment includes, but is not limited to, equipment number and running time, and can be managed through an equipment management system.

[0051] In this embodiment, a binocular machine vision camera is used, featuring a high resolution of ≥5 megapixels and a high frame rate of ≥30fps. It is equipped with a telecentric lens and a dedicated light source (such as a ring light source or a bar light source) to ensure clear, distortion-free image data of the battery. The camera is deployed behind the rolling station and above the inspection station to perform real-time inspection of the appearance and dimensions of the rolled battery. At the rolling station, the machine vision camera primarily detects defects such as scratches, dents, and contamination on the battery surface. At the inspection station, the machine vision camera precisely measures the battery's thickness, width, length, and other dimensional parameters.

[0052] The control layer is used to detect defects and dimensions of the battery based on image data, obtain detection results, and generate control commands based on equipment status data, process parameter data, and detection results.

[0053] In this embodiment, the control layer mainly consists of a programmable logic controller (PLC) and an image processor. The PLC is used to generate corresponding control instructions, and the image processor is used to process image data to perform defect and size detection on the battery.

[0054] In this embodiment, to avoid the adverse effects of lighting conditions on battery defect and size detection, the light source is optimized as follows: a ring light source is used as the light source for battery appearance defect detection, which can provide uniform illumination and effectively suppress battery surface reflection; a strip light source is used as the light source for battery size detection, which can highlight the battery edge contour; therefore, the detection accuracy is improved through the above optimization of the light source.

[0055] In this embodiment, the control layer performs defect detection on the battery, including the following steps:

[0056] First, the image data is preprocessed to obtain the target image. The preprocessing in this embodiment includes, but is not limited to, denoising, filtering, grayscale transformation, etc., to eliminate noise interference, enhance image contrast, and improve the accuracy of subsequent feature extraction.

[0057] Then, feature extraction is performed on the target image to obtain feature information. In this embodiment, algorithms such as edge detection, contour extraction, and threshold segmentation can be used to extract feature information such as the edge contour and defect area of ​​the battery.

[0058] Finally, the feature information is input into a pre-built defect identification model for defect prediction to obtain the defect type of the battery, and the defect type of the battery is used as the detection result. The defect identification model in this embodiment can be a neural network, support vector machine, etc. The defect types in this embodiment include, but are not limited to, scratches, dents, contamination, etc. When a defect is detected in the battery, the system marks the battery and generates a corresponding sorting instruction to sort and remove the battery.

[0059] The defect recognition model in this embodiment can construct a training dataset by collecting a large amount of battery defect image data, and then train and optimize the neural network and support vector machine. The trained model can accurately identify various defect types such as scratches, dents, and contamination, and the detection accuracy is improved compared with traditional detection algorithms based on manual features.

[0060] Since the machine vision camera in this embodiment is a binocular camera, the image data consists of two viewpoint images captured by the binocular camera. Therefore, the step of the control layer to detect the size of the battery includes:

[0061] The first step is to calibrate the stereo camera based on a preset algorithm and obtain the calibration results.

[0062] The preset algorithm in this embodiment is Zhang's calibration method. Therefore, the steps for calibrating the binocular camera based on the preset algorithm include:

[0063] A1: Using Zhang's calibration method, the two cameras of a stereo camera are individually calibrated to obtain the intrinsic and extrinsic parameters of each camera. The intrinsic parameters include focal length, principal point coordinates, and distortion parameters. The extrinsic parameters include rotation matrix and translation vector. In this embodiment, Zhang's calibration method establishes the transformation relationship between the image coordinate system and the world coordinate system by capturing a series of calibration board images, thereby eliminating the influence of camera lens distortion on the image and achieving accurate image measurement and 3D reconstruction. Its specific principle is as follows:

[0064] Calibration board model: The calibration board usually adopts a black and white checkerboard pattern, and its geometric position and size are known; the coordinates of the corner points of the checkerboard can be calculated accurately and are used as known points;

[0065] Image acquisition: Multiple images of the calibration board from different angles and positions are captured using a camera. The images contain the coordinates of the checkerboard corner points, which are used as the points to be determined.

[0066] Image processing: The acquired images are preprocessed, including denoising and grayscale conversion, and then corner detection algorithms (such as Harris corner detection) are used to extract the checkerboard corners in the image;

[0067] Distortion model: Establish a camera lens distortion model to describe the difference between actual imaging and ideal imaging. The distortion model can be a radial distortion model or a tangential distortion model.

[0068] Calibration calculation: Based on the calibration board model and distortion model, establish the transformation relationship between the image coordinate system and the world coordinate system, and then use optimization algorithms such as the least squares method to solve the camera intrinsic parameters (focal length, principal point coordinates) and extrinsic parameters (rotation matrix, translation vector).

[0069] A2: Construct coordinate systems for the two cameras and transform them to the same world coordinate system. This can be achieved by setting a common reference point or by matching the common view of the two cameras.

[0070] A3: In the world coordinate system, the relative position and orientation of the two cameras are determined based on the rotation matrix and translation vector between them.

[0071] The second step is to use the calibration results to perform stereo correction on the two viewpoint images to obtain corrected images from the two viewpoints.

[0072] In this embodiment, the calibration results are used to perform stereo correction on the left and right camera images so that the imaging planes are aligned in a coplanar manner. That is, the left and right camera image planes are on the same plane, and when the same point is projected onto the two camera image planes, it should be in the same row of the two pixel coordinate systems.

[0073] The third step is to perform stereo matching on the corrected images from the two perspectives to obtain a disparity map. In this embodiment, each pixel value of the disparity map represents the disparity value of that pixel, that is, the positional difference of the pixel in the left and right images.

[0074] In this embodiment, the SGM (Semi-Global Matching) algorithm can be used to perform stereo matching on the corrected images from two viewpoints. The processing steps of the SGM algorithm are as follows:

[0075] First, cost calculation: The cost of each pixel between image pairs is calculated to reflect the degree of pixel matching. Census transform is usually used to calculate the cost. Census transform selects a pixel as a reference point and constructs a 3*3 or 5*5 pixel window with the reference point as the center. Then, from top to bottom and from left to right, the gray values ​​of the pixels are compared with the gray values ​​of the reference point. If the gray value of the reference point is greater than or equal to the gray value of the pixel, it is marked as "1" and less than "0". The comparison results are then encoded into a set of binary codes. Finally, the binary codes of the left and right cameras are substituted into the Hamming distance formula to obtain the cost based on Census transform.

[0076] Then, cost aggregation: using algorithms such as beam search or graph cut, the local costs are aggregated into global costs to obtain the disparity map.

[0077] Finally, disparity optimization: the disparity map is optimized, for example, by smoothing, eliminating noise and outliers.

[0078] In this embodiment, since the Census transform selects the center point of the pixel window as the reference point, when the pixel at this center point is affected by other environmental noise and its grayscale value changes significantly, it will cause a serious deviation in the cost calculation result, adversely affecting the stereo matching result, and thus leading to excessive errors in battery size detection. Furthermore, when the grayscale value of pixels at the same position in the left and right camera windows differs too much from the reference point, the mismatch rate of the Census transform will also increase, similarly reducing the accuracy of size detection. To solve the above problems, this embodiment uses the following steps for stereo matching:

[0079] B1. Select any pixel in the corrected image from two perspectives as the target pixel and construct a pixel window for the target pixel, wherein the center pixel of the pixel window is the target pixel. In this embodiment, the pixel window is a 3*3 or 5*5 pixel window. For pixels on the edge in the corrected image, the constructed pixel window needs to fill the blank pixels. For example, the average value of all existing pixels in the pixel window can be calculated as the fill value.

[0080] B2. Determine the reference pixel point of the pixel window based on a preset selection rule. The preset selection rule is as follows: calculate the average pixel value within the pixel window, calculate the absolute pixel difference between the average pixel value and the target pixel value; determine if the absolute pixel difference is less than a preset value. If so, use the target pixel value as the reference pixel value; otherwise, use the average pixel value as the reference pixel value. This selection rule effectively avoids selecting a pixel value with a significant abrupt change as the reference pixel, thereby improving the accuracy of subsequent size measurements.

[0081] B3. Based on the reference pixel and each pixel within the pixel window, determine the matching cost of the target pixel; the matching cost in this embodiment is the same as the cost calculation steps in the SGM algorithm.

[0082] B4. Based on a preset correction rule, the matching cost of the target pixel is corrected to obtain the corrected matching cost of the target pixel; each pixel in the corrected image from both views is traversed to obtain the corrected matching cost corresponding to each pixel; the preset correction rule in this embodiment is:

[0083] Calculate the rate of change between each pixel in each pixel window and the reference pixel. The rate of change is calculated by subtracting the pixel value of the pixel from the pixel value of the reference pixel to obtain the pixel difference, taking the absolute value of the pixel difference to obtain the absolute pixel difference, and then dividing the absolute pixel difference by the pixel value of the reference pixel to obtain the rate of change between the pixel and the reference pixel.

[0084] For pixel windows corresponding to pixels at the same position in the corrected images from two perspectives, the change rates of each pixel in the two pixel windows are subtracted to obtain the change difference. All change differences are summed to obtain the cost correction value. The cost correction value is used to correct the matching cost corresponding to the pixel window, that is, the cost correction value is added to the matching cost corresponding to the pixel window to obtain the corrected matching cost of the target pixel in the pixel window.

[0085] This embodiment can effectively reduce the false matching rate and improve the robustness of stereo matching operations on the corrected image under noise by correcting the matching cost of the target pixel, thereby improving the matching effect.

[0086] B5. Aggregate the corrected matching costs corresponding to each pixel to obtain a disparity map. The aggregation operation steps in this embodiment are the same as the cost aggregation in the SGM algorithm.

[0087] The fourth step is to obtain the focal length and baseline length of the binocular camera, and determine the depth map based on the focal length and baseline length of the binocular camera and the disparity map. The depth map is used to characterize the three-dimensional coordinates of each pixel in the two-dimensional image.

[0088] In this embodiment, the baseline length can be obtained by directly measuring the distance between the centers of the left and right camera lenses using a ruler or similar tool; however, this method introduces measurement errors, leading to significant errors in the calculated depth map. Therefore, to improve the accuracy of baseline length acquisition, this embodiment preferably acquires the baseline length using the following method: Single-target calibration is performed on each of the two cameras of the binocular camera, allowing the calculation of the rotation and translation matrices between the left and right cameras. Using these matrices, the distance between the optical centers of the left and right cameras can be calculated, and this distance is then used as the baseline length.

[0089] In this embodiment, depth can be calculated using the principle of triangulation based on the disparity map and camera parameters. The formula for calculating depth is as follows:

[0090] Z = fB / D;

[0091] In the formula, Z is the depth, f is the focal length, D is the disparity value in the disparity map, and B is the baseline length.

[0092] After calculating the depth, the three-dimensional coordinates (X,Y,Z) can be calculated by combining the pixel coordinates (x,y) and depth (Z) of the two-dimensional image, as follows:

[0093] X=(xc x )*Z / f;

[0094] Y=(yc y )*Z / f;

[0095] In the formula, c x and c y These are the x and y coordinates of the camera's optical center, respectively.

[0096] The calculated depth information is presented in the form of an image, which is called a depth map. Each pixel value in the depth map represents the depth information corresponding to that pixel.

[0097] The depth map in this embodiment has the advantages of being smoother and more stable, which can avoid distance inaccuracies and improve the accuracy of size measurement.

[0098] The fifth step is to determine the battery size information based on the depth map. The size information includes at least the battery's length, width, and thickness, and the battery size information is used as the detection result.

[0099] In this embodiment, two points that can measure the size of the object being measured can be selected in the depth map, such as two pixels on the edge of the object. The geometric distance between these two points is calculated using the Euclidean distance formula to obtain the size information of the object.

[0100] The execution layer is used to respond to control commands generated by the control layer and perform battery production operations, including rolling operations, transfer operations, and sorting operations.

[0101] In this embodiment, the execution layer includes a rolling device, a conveying device, and a sorting device, all of which are controlled by a programmable logic controller (PLC). Therefore, the control instructions include a rolling instruction, a conveying instruction, and a sorting instruction. The rolling device is used to respond to the rolling instruction and perform a rolling operation on the battery. The conveying device is used to respond to the conveying instruction and perform a conveying operation on the battery. The sorting device is used to respond to the sorting instruction and perform a sorting operation on the battery.

[0102] In this embodiment, the equipment status data and process parameter data can be used as feedback data for the programmable logic controller (PLC). The PLC can generate control commands using a PID algorithm. The PLC in this embodiment is also used for synchronous control of the production line. Specifically, to ensure coordinated operation between the rolling equipment, the conveying equipment, and the testing equipment, the PLC adopts a master-slave control mode to achieve synchronous control of the production line. The rolling equipment is the main power source, and the conveying and testing equipment are slave devices. The encoder acquires the running speed and position information of the rolling equipment in real time. The slave devices make synchronous adjustments based on the status information of the master equipment to ensure that the batteries accurately enter the testing and rolling stations during transmission, avoiding accumulation or jamming.

[0103] The management layer is used to store, trace, manage, and visualize the battery production data collected by the perception layer and the detection results detected by the control layer.

[0104] In this embodiment, the management layer can be a cloud server. The battery production data collected by the perception layer and the detection results detected by the control layer are uploaded to the cloud server via industrial Ethernet. The cloud server stores, traces, manages, and visualizes this data.

[0105] The cloud server in this embodiment adopts a distributed storage architecture, storing image data, process parameter data, equipment status data, etc. in different databases to improve data query and analysis efficiency. The data storage period is [X] years, which meets the needs of production traceability and quality analysis.

[0106] In this embodiment, the step of the management layer to trace and manage the battery production data collected by the sensing layer and the detection results detected by the control layer includes:

[0107] First, a unique identifier is pre-assigned to each battery. During battery production, a unique batch identifier (such as a QR code or barcode) is assigned to each batch of batteries, and this batch identifier is printed on the battery casing or packaging using an inkjet printer. A machine vision camera identifies the batch identifier during inspection, binding the visual inspection data of that batch of batteries with the batch information.

[0108] Then, based on the production timeline, the unique identifier of each battery is associated with the corresponding equipment status data, process parameter data, and test results to obtain associated data; process parameters such as rolling pressure, temperature, and speed are collected in real time by sensors and transmitted to the cloud server via industrial Ethernet; the cloud server associates and binds the process parameters with the corresponding battery batch and visual inspection data according to the production timeline.

[0109] Finally, the operator's identity information is obtained and bound to related data to obtain traceability data; the operating status information (such as equipment number, running time, fault records, etc.) of rolling equipment, testing equipment, etc. is obtained through the equipment management system, and the operator's identity information (such as employee number, name, operation time, etc.) is obtained through the personnel management system, and this information is bound to production data to achieve full traceability of the production process.

[0110] Therefore, by entering the battery batch number or product serial number, users can query the production process data for that batch of batteries, including rolling process parameters, visual inspection results, equipment operating status, and operator information. The system also features data statistical analysis capabilities, enabling it to statistically analyze quality data during the production process, generate quality reports and trend charts, and help managers promptly identify potential problems in the production process and optimize production processes and quality management strategies.

[0111] Therefore, the integrated battery production system in this embodiment realizes real-time detection, data binding, and intelligent traceability of the production process, improving production efficiency, quality stability, and traceability. By detecting defects and dimensions of the batteries, surface defects and dimensional deviations can be detected in a timely manner, and defective products can be removed by automated sorting equipment, preventing defective products from flowing into the next process.

[0112] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A battery integration production system based on visual inspection, characterized by, The system comprises: a perception layer deployed on a battery production line for collecting production data of the battery, the production data comprising image data, equipment state data and process parameter data; a control layer for detecting defects and sizes of the battery according to the image data to obtain detection results, and generating control instructions according to the equipment state data, the process parameter data and the detection results; an execution layer for executing production operations of the battery in response to the control instructions generated by the control layer, the production operations comprising rolling operations, transmission operations and sorting operations; a management layer for storing, traceability management and visual display of the production data of the battery collected by the perception layer and the detection results detected by the control layer.

2. The battery integration production system based on visual inspection according to claim 1, wherein, The execution layer comprises rolling equipment, transmission equipment and sorting equipment, and the control instructions comprise rolling instructions, transmission instructions and sorting instructions. The rolling equipment is configured to execute the rolling operations of the battery in response to the rolling instructions. The transmission equipment is configured to execute the transmission operations of the battery in response to the transmission instructions. The sorting equipment is configured to execute the sorting operations of the battery in response to the sorting instructions.

3. The battery integration production system based on visual inspection according to claim 1, characterized by, The management layer performs the traceability management of the production data of the battery collected by the perception layer and the detection results detected by the control layer, comprising: obtaining a unique identifier pre-assigned to each battery; associating and binding the unique identifier of each battery with the corresponding equipment state data, process parameter data and detection results based on a production timeline to obtain association data; obtaining identity information of an operator and binding the identity information with the association data to obtain traceability data.

4. The battery integration production system based on visual inspection according to claim 1, characterized by, The control layer performs the defect detection of the battery, comprising: preprocessing the image data to obtain a target image; extracting features from the target image to obtain feature information; inputting the feature information into a pre-constructed defect recognition model for defect prediction to obtain a defect type of the battery, and taking the defect type of the battery as the detection result.

5. The battery integration production system based on visual inspection according to claim 4, characterized by, The image data is two-view images collected by a binocular camera, and the control layer performs the size detection of the battery, comprising: calibrating the binocular camera based on a preset algorithm to obtain a calibration result; performing stereo correction on the two-view images based on the calibration result to obtain corrected images of the two views; performing stereo matching on the corrected images of the two views to obtain a disparity map; obtaining a focal length and a baseline length of the binocular camera, and determining a depth map based on the focal length and the baseline length of the binocular camera and the disparity map, the depth map being used to represent three-dimensional coordinates of each pixel point in a two-dimensional image; determining size information of the battery based on the depth map, the size information at least comprising length, width and thickness of the battery, and taking the size information of the battery as the detection result.

6. The battery integration production system based on visual inspection according to claim 5, wherein, The preset algorithm is Zhang's calibration method, and the calibration of the binocular camera based on the preset algorithm comprises: performing single target calibration on the two cameras of the binocular camera respectively by Zhang's calibration method to obtain intrinsic and extrinsic parameters of each camera, the intrinsic parameters comprising focal length, principal point coordinates and distortion parameters, and the extrinsic parameters comprising a rotation matrix and a translation vector; Coordinate systems of the two cameras are constructed, and the coordinate systems of the two cameras are converted to the same world coordinate system; In the world coordinate system, the relative position and attitude between the two cameras are determined according to the rotation matrix and the translation vector between the two cameras.

7. The battery integration production system based on visual inspection according to claim 5, characterized by, The step of performing stereo matching on the corrected images of the two views includes: selecting an arbitrary pixel point in the corrected images of the two views as a target pixel point, and constructing a pixel window of the target pixel point, the center pixel of the pixel window being the target pixel point; determining a reference pixel point of the pixel window based on a preset selection rule; determining a matching cost of the target pixel point based on the reference pixel point and each pixel point in the pixel window; correcting the matching cost of the target pixel point based on a preset correction rule to obtain a corrected matching cost of the target pixel point; and traversing each pixel point in the corrected images of the two views to obtain a corrected matching cost corresponding to each pixel point; aggregating the corrected matching costs corresponding to each pixel point to obtain a disparity map.

8. The battery integration production system based on visual inspection according to claim 7, characterized by, The preset selection rule is: calculating an average pixel value in the pixel window, calculating an absolute pixel difference between the average pixel value and the pixel value of the target pixel point; determining whether the absolute pixel difference is less than a preset value, if yes, taking the pixel value of the target pixel point as the pixel value of the reference pixel point; if not, taking the average pixel value as the pixel value of the reference pixel point.

9. The battery integration production system based on visual inspection according to claim 7, wherein, The preset correction rule is: calculating a change rate between each pixel point in each pixel window and the reference pixel point; for the pixel windows corresponding to the pixel points at the same position in the corrected images of the two views, subtracting the change rates corresponding to the pixel points in the two pixel windows to obtain a change difference value; summing all the change difference values to obtain a cost correction value, and using the cost correction value to correct the matching cost corresponding to the pixel window to obtain the corrected matching cost of the target pixel point in the pixel window.

Images