A battery detection method, device and equipment

Abstract

The present invention provides a battery detection method, device and equipment. The method comprises: obtaining first image data to be detected, second data to be detected and third data to be detected of a lithium battery; performing surface detection on the lithium battery according to the first image data to be detected and a pre-trained target area recognition model to obtain a surface detection result; performing internal detection on the lithium battery according to the second data to be detected to obtain an internal detection result; performing performance detection on the lithium battery according to the third data to be detected and a preset detection threshold to obtain a performance detection result; filling a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain an automatic detection result of the lithium battery. The present invention can not only realize automatic detection of lithium batteries by processing the acquired data, but also improve the comprehensiveness and accuracy of lithium battery detection.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery detection, and also to a battery detection method, device and equipment. Background Art

[0002] With the continuous development of the information society and the new energy industry, lithium-ion batteries have been widely used and developed in many fields such as consumer electronics, automobiles, communications, aerospace and navigation due to their advantages such as high output voltage, large specific energy, long cycle life, good safety performance, low self-discharge, and fast charging speed. When soft-pack lithium batteries are automatically packaged, online testing is required at key process nodes to improve the production yield of the product. The existing testing methods for lithium batteries generally test the interior or surface of the lithium battery separately, which requires multiple testing processes to detect whether there are defects inside or on the surface of the lithium battery. The testing method is relatively cumbersome. On the one hand, it requires manual participation in identification and testing, and on the other hand, it requires manual output of the test report, and the degree of automation is low. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a battery detection method, device and equipment to improve the comprehensiveness and accuracy of lithium battery detection.

[0004] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0005] A first aspect of the present invention provides a battery detection method, comprising:

[0006] Acquire first image data to be detected, second data to be detected, and third data to be detected of the lithium battery;

[0007] Performing surface detection on the lithium battery according to the first to-be-detected image data and a pre-trained target area recognition model to obtain a surface detection result;

[0008] Performing an internal test on the lithium battery according to the second data to be tested to obtain an internal test result;

[0009] Performing a performance test on the lithium battery according to the third data to be tested and a preset test threshold to obtain a performance test result;

[0010] The preset automatic detection result template is filled according to the surface detection result, the internal detection result and the performance detection result to obtain the automatic detection result of the lithium battery.

[0011] Optionally, the first image data to be detected is a surface image of a lithium battery collected by a camera device, and the target area recognition model is obtained by training a first preset network model using a battery feature image and its corresponding label; performing surface detection on the lithium battery according to the first image data to be detected and the pre-trained target area recognition model to obtain a surface detection result includes:

[0012] Correcting the surface image of the lithium battery to obtain a corrected surface image;

[0013] Performing region extraction on the corrected surface image to obtain a region to be processed;

[0014] The lithium battery is subjected to surface detection according to the area to be processed and a pre-trained target area recognition model to obtain a surface detection result.

[0015] Optionally, performing surface detection on the lithium battery according to the area to be processed and a pre-trained target area recognition model to obtain a surface detection result includes:

[0016] Inputting the area to be processed into a pre-trained target area recognition model to perform target area recognition to obtain a target area;

[0017] Performing grayscale adjustment on the target area to obtain an adjusted target area;

[0018] Determining a defect image according to a preset threshold and the adjusted target area;

[0019] Performing fusion processing on the defect image to obtain a fused defect image;

[0020] According to the fused defect image and the trained defect recognition model, the lithium battery is subjected to surface inspection to obtain a surface inspection result; the defect recognition model is obtained by training a second preset network model using the surface defect image of the lithium battery.

[0021] Optionally, performing surface inspection on the lithium battery according to the fused defect image and the trained defect recognition model to obtain a surface inspection result includes:

[0022] Inputting the fused defect image into the input layer of the defect recognition model for processing to obtain a first output result;

[0023] Inputting the first output result into a mapping layer of a defect recognition model for processing to obtain a second output result;

[0024] Inputting the second output result into the separation layer of the defect recognition model for processing to obtain a third output result;

[0025] The third output result is input into the output layer of the defect recognition model for processing to obtain a surface detection result.

[0026] Optionally, the second data to be detected is an ultrasonic signal of a lithium battery and an internal image of the lithium battery collected by an air-coupled ultrasonic detection system; performing internal detection on the lithium battery according to the second data to be detected to obtain an internal detection result includes:

[0027] Extracting characteristic parameters of the ultrasonic signal of the lithium battery to obtain characteristic parameters;

[0028] An internal inspection of the lithium battery is performed according to the characteristic parameters and the internal image of the lithium battery to obtain an internal inspection result.

[0029] Optionally, the third data to be detected is voltage data of the lithium battery collected by a voltage collection device and battery capacity data of the lithium battery collected by a battery capacity detection device, and the preset detection threshold includes a preset voltage detection threshold and a preset battery capacity detection threshold; and the lithium battery is subjected to performance detection according to the third data to be detected and the preset detection threshold to obtain a performance detection result, including:

[0030] Performing a performance test on the lithium battery according to the voltage data of the lithium battery and a preset voltage detection threshold to obtain a voltage detection result;

[0031] Performing a performance test on the lithium battery according to the battery capacity data of the lithium battery and a preset battery capacity test threshold to obtain a battery capacity test result;

[0032] A performance test result is obtained according to the voltage test result and the battery capacity test result.

[0033] Optionally, filling a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain the automatic detection result of the lithium battery includes:

[0034] Get the preset automatic detection result template;

[0035] Extracting features from the surface detection result, the internal detection result, and the performance detection result according to the preset automatic detection result template to obtain a feature set;

[0036] The preset automatic detection result template is filled according to the feature set to obtain the automatic detection result of the lithium battery.

[0037] A second aspect of the present invention provides a battery detection device, comprising:

[0038] An acquisition module, used to acquire first image data to be detected, second data to be detected and third data to be detected of the lithium battery;

[0039] A first detection module, used to perform surface detection on the lithium battery according to the first to-be-detected image data and a pre-trained target area recognition model to obtain a surface detection result;

[0040] A second detection module, used for performing an internal detection on the lithium battery according to the second to-be-detected data to obtain an internal detection result;

[0041] A third detection module, used to perform a performance test on the lithium battery according to the third to-be-detected data and a preset detection threshold, to obtain a performance test result;

[0042] The processing module is used to fill in a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain the automatic detection result of the lithium battery.

[0043] According to a third aspect of the present invention, a computing device is provided, comprising: a processor and a memory storing a computer program, wherein when the computer program is executed by the processor, the method according to the first aspect is executed.

[0044] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions, which, when executed on a computer, causes the computer to execute the method described in the first aspect.

[0045] The above solution of the present invention includes at least the following beneficial effects:

[0046] The above scheme of the present invention realizes surface detection, internal detection and performance detection of lithium batteries by acquiring first image data to be detected, second data to be detected and third data to be detected of lithium batteries, and automatically processes these data by using a pre-trained target area recognition model and a preset detection threshold, etc., and fills in the obtained surface detection results, internal detection results and performance detection results using a preset automatic detection result template to obtain the automatic detection result of the lithium battery. The automatic detection result template can be used to realize automatic generation of detection results, effectively reducing manual participation, which can not only realize automatic detection of lithium batteries, but also improve the comprehensiveness and accuracy of lithium battery detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a flowchart of a battery detection method in an embodiment of the present invention;

[0048] Figure 2 is a schematic diagram of the internal structure of the first preset network model in an embodiment of the present invention;

[0049] Figure 3 is a schematic diagram of the internal structure of a second preset network model in an embodiment of the present invention;

[0050] Figure 4 It is a structural schematic diagram of a battery detection device in an embodiment of the present invention.

[0051] Explanation of the reference numerals: 11 - input layer, 12 - hidden layer, 13 - output layer, 21 - hyperplane, 22 - defect information, 23 - non-defect information. DETAILED DESCRIPTION

[0052] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0053] like Figure 1 As shown, an embodiment of the present invention provides a battery detection method, comprising the following steps:

[0054] Step 101, obtaining first image data to be detected, second data to be detected, and third data to be detected of a lithium battery;

[0055] Step 102, performing surface detection on the lithium battery according to the first to-be-detected image data and a pre-trained target area recognition model to obtain a surface detection result;

[0056] Step 103, performing an internal test on the lithium battery according to the second data to be tested, and obtaining an internal test result;

[0057] Step 104, performing a performance test on the lithium battery according to the third to-be-tested data and a preset test threshold to obtain a performance test result;

[0058] Step 105 , filling a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain the automatic detection result of the lithium battery.

[0059] Specifically, the first image data to be detected may include a surface image of a lithium battery captured by a camera device, which is used for surface detection of the lithium battery, i.e., detecting whether the appearance or surface of the lithium battery has defects such as concave spots, dents, pits, uneven edges, and flatulence; the second data to be detected is the ultrasonic signal of the lithium battery and the internal image of the lithium battery collected by an air-coupled ultrasonic detection system, which is used for internal detection of the lithium battery, i.e., detecting whether there are defects such as bubbles and stratification inside the lithium battery; the third data to be detected is the voltage data of the lithium battery collected by a voltage acquisition device and the battery capacity data of the lithium battery collected by a battery capacity detection device, which is used for performance detection of the lithium battery, i.e., detecting whether the voltage, battery capacity, etc. of the lithium battery meet the expected requirements.

[0060] The battery detection method of the embodiment of the present invention obtains the first image data to be detected, the second data to be detected and the third data to be detected of the lithium battery, and automatically processes these data using a pre-trained target area recognition model and a preset detection threshold, etc., to achieve surface detection, internal detection and performance detection of the lithium battery. According to the obtained surface detection results, internal detection results and performance detection results, a preset automatic detection result template is used to fill in and obtain the automatic detection result of the lithium battery. The automatic detection result template can be used to achieve automatic matching of the detection results, effectively reducing manual participation, which can not only achieve automatic detection of lithium batteries, but also improve the comprehensiveness and accuracy of lithium battery detection.

[0061] It should be noted that although step 102, step 103 and step 104 are marked above, these three steps can be adjusted according to actual conditions, such as step 103, step 102 and step 104 or step 104, step 102 and step 103, etc. This embodiment is only for illustrative purposes and is not intended to limit the present invention.

[0062] In an optional embodiment of the present invention, the first image data to be detected is a surface image of a lithium battery collected by a camera device; the target area recognition model is obtained by training a first preset network model using a battery feature image and its corresponding label; step 102 includes:

[0063] Step 1021, correcting the surface image of the lithium battery to obtain a corrected surface image;

[0064] Specifically, the production process of lithium batteries goes through multiple steps. Starting from the battery pole, there may be lithium battery surface defects such as dents, dents, pits, uneven edges, bloating, and welding defects such as tinning, tin slag and cold soldering. These defects can be detected through the appearance image or external surface image of the lithium battery. Therefore, the first image data to be detected may include the surface image of the lithium battery captured by the camera equipment, which is used for surface detection of the lithium battery.

[0065] Specifically, since the collected image may be unclear, inverted, or other problems, the surface image of the lithium battery is first corrected, and the correction method may be at least one of the following methods:

[0066] Geometric correction, adjusting the geometric shape of the image, such as rotation, scaling, translation, etc., to ensure that the geometric characteristics of the image match the preset standard image;

[0067] Radiation correction, adjusting the brightness and contrast of the image to ensure that the radiation characteristics of the image are consistent with the preset standard image;

[0068] Color correction adjusts the color of the image to conform to the color space of the preset standard image.

[0069] In addition, after the correction process, the corrected surface image can also be preprocessed. The preprocessing method can include noise processing to reduce the noise in the image and improve the clarity of the image; or enhancement processing to enhance certain features of the image, such as edges, textures, etc., to improve the visual effect of the image.

[0070] Step 1022, performing region extraction on the corrected surface image to obtain a region to be processed;

[0071] Specifically, the corrected surface image is firstly gray-scaled and converted into a binary image so as to more clearly identify the edge, and then the edge in the binary image is identified by edge detection.

[0072] Then, according to the edges in the binary image, the contours in the binary image are searched. According to the shape, size, direction and other attributes of the contour, the contours that meet the preset requirements are screened. For example, the area of ​​the contour, the aspect ratio of the circumscribed rectangle and other features can be used for screening. By calculating the circumscribed rectangle of the contour, check whether its direction is close to horizontal. This can be achieved by comparing the width and height of the circumscribed rectangle. If the width is much greater than the height, the contour can be considered to be placed horizontally. According to the screened contours, the target area is determined. Usually, the target area can be set as the circumscribed rectangle of the contour. Finally, according to the coordinates of the target area (x, y coordinates of the upper left corner and width, height), the target area is cropped out from the corrected surface image using array slicing to obtain the area to be processed.

[0073] Step 1023, performing surface detection on the lithium battery according to the area to be processed and a pre-trained target area recognition model to obtain a surface detection result; the target area recognition model is obtained by training a first preset network model using a battery feature image and its corresponding label.

[0074] In an optional embodiment of the present invention, step 1023 includes:

[0075] Step 10231, inputting the area to be processed into a pre-trained target area recognition model to perform target area recognition to obtain a target area;

[0076] Specifically, the first preset network model is trained using battery feature images (including images of the appearance of lithium batteries) and their corresponding labels (such as pole pieces, lithium battery bodies, etc.) as data sets. By adjusting the weights and biases of the first preset network model, the first preset network model can accurately identify the parts of lithium batteries in different images, thereby obtaining a target area recognition model.

[0077] like Figure 2 As shown, the first preset network model includes an input layer 11, a hidden layer 12 and an output layer 13. Before training the first preset network model, the neuron nodes of the input layer 11, the hidden layer 12 and the output layer 13 need to be set. In this embodiment, the input data includes the battery feature image and the corresponding label, and the number of neuron nodes of the input layer 11 is set to 2; the output data includes the names and corresponding positions of the various parts of the lithium battery, and the number of neuron nodes of the output layer 13 is set to 2; the number of neuron nodes in the hidden layer 12 is uncertain, but the more the number, the more significant the nonlinearity of the model, and thus the robustness of the model. Generally, the number of nodes of the i-th layer of neurons is 1 to 1.5 times the number of nodes of the i-1 layer. The process of training the first preset network model is to continuously modify the weight vector and bias to make it as close to the real model as possible, so that the recognition effect of the target area recognition model is finally obtained.

[0078] The training process of the first preset network model is:

[0079] Collect images containing the appearance of lithium batteries (such as images of lithium battery appearance containing pole pieces, images of lithium battery appearance containing lithium battery bodies) and corresponding labels (such as pole pieces, lithium battery bodies, etc.) as training samples.

[0080] Assign random values ​​to all weight vectors and biases in the first preset network model and define the loss function , where L is the loss function, is the output result of the model, and y is the true result.

[0081] The training samples are input into the first preset network model for training until the loss function reaches a preset minimum value, and the training is stopped to obtain the target weight vector and the target bias.

[0082] At this point, all parameters in the first preset network model (number of nodes in the input layer, output layer, hidden layer, target weight vector, target bias) are known, so the first preset network model has been trained as a target region recognition model. It is only necessary to input the area to be processed from the input layer of the target region recognition model, and perform calculations in the model network along the direction of data flow until the data is transferred to the output layer and output, and then a target region recognition is completed and the target region is obtained.

[0083] The area to be processed is input into the input layer of the pre-trained target area recognition model for format conversion, and the area to be processed is converted into a format that can be processed by the target area recognition model (such as converted into a numerical format or a vector format) to obtain a conversion result; then the first recognition result is input into at least one hidden layer of the target area recognition model, and feature extraction is performed on the first recognition result to obtain a feature extraction result; finally, the feature extraction result is input into the output layer of the target area recognition model, and target area recognition and classification are performed to obtain a target area, which includes the names of various parts of the lithium battery (such as pole pieces, lithium battery body, etc.) and corresponding positions.

[0084] Step 10232, grayscale adjustment is performed on the target area to obtain an adjusted target area;

[0085] Specifically, the grayscale value of each pixel point inside the target area is calculated, and its distribution is counted. Then, according to the distribution of the grayscale value, the grayscale value of the target area is adjusted to make it relatively uniform.

[0086] Step 10233, determining a defect image according to a preset threshold and the adjusted target area;

[0087] Specifically, the adjusted target area needs to be divided into multiple sub-images (i.e., sub-areas), which can be overlapping or non-overlapping, and are selected according to actual conditions; an adaptive grayscale threshold is calculated for each sub-image, and this threshold is usually determined based on the statistical characteristics of the grayscale values ​​of pixels in the sub-image (such as mean, median, standard deviation, etc.); the grayscale value of each pixel in the sub-image is compared with the corresponding adaptive threshold, and if the grayscale value of a pixel exceeds a preset threshold range (usually higher than the threshold, but in some cases it may be lower than the threshold), the pixel is considered to belong to the defective area; if the proportion of defective pixels in a sub-image exceeds a preset value (such as 5%), or a sub-image contains a preset number (such as 10) of defective pixels, the sub-image is set as a defective image.

[0088] Step 10234, performing fusion processing on the defect image to obtain a fused defect image;

[0089] Specifically, the purpose of fusing defect images is to merge multiple defect images that may contain defect information into a more complete image, so as to more accurately identify and analyze defect areas. The fusion process can be based on the following principles: Overlapping area processing: If there is overlap between defect images, the pixel values ​​of the overlapping area need to be processed by weighted averaging or maximum value selection to ensure that the fused image maintains a smooth transition in the overlapping part. Feature retention: During the fusion process, the key features in each defect image should be retained as much as possible, especially the features related to the defect, such as shape, size, color, etc. Global consistency: The fused image should maintain global consistency and coherence to avoid obvious splicing marks or distortion.

[0090] Here, the fused defective image needs to be filtered out, that is, the non-defective area is filtered out to reduce noise and interference and improve the accuracy and efficiency of defect detection. This can be achieved by the following steps: Threshold segmentation: The fused image is segmented using an adaptive threshold or a global threshold to divide the image into defective areas and non-defective areas. The adaptive threshold or the global threshold can be set according to the actual situation. Morphological operation: Morphological operations (such as corrosion, expansion, opening operation, closing operation, etc.) are performed on the segmented image to further remove noise and refine the defective area. These operations can be based on the principles of mathematical morphology and can be implemented by defining structural elements to process specific shapes and structures in the image. Connectivity analysis: Connectivity analysis is performed on the image after morphological operation to filter out isolated noise points or small areas, and only defective areas with a certain size and connectivity are retained. Through filtering, a filtered defective image is obtained. Subsequently, the filtered defective image can be input into a trained defect recognition model to perform surface inspection on lithium batteries.

[0091] Step 10235, performing surface inspection on the lithium battery according to the fused defect image and the pre-trained defect recognition model to obtain a surface inspection result; the defect recognition model is obtained by training a second preset network model using the lithium battery surface defect image.

[0092] Here, the labeled lithium battery surface defect images are used as training samples to train the second preset network model. During the training process, the model will learn how to identify defects in the image based on the extracted features and output information such as the location and type of the defects.

[0093] like Figure 3 The figure shows the internal structure of the second preset network model. The purpose of training the second preset network model is to find an optimal hyperplane 21 and obtain a defect recognition model to classify defect information 22 and non-defect information 23 in the fused defect image and obtain the location and type of the defect. Figure 3 The horizontal axis is the x-axis and the vertical axis is the y-axis.

[0094] The training process of the second preset network model is:

[0095] Collect lithium battery surface defect images and perform preprocessing such as data cleaning, feature selection and data standardization to improve data quality and make the defect recognition model trained in the end more effective. The preprocessed battery surface defect images are divided into training samples and test samples.

[0096] Select kernel functions and hyperparameters. The kernel function is used to map data from the original space to a high-dimensional feature space so as to perform linear classification in the high-dimensional feature space and improve the processing efficiency of the model. The kernel function used in this embodiment is: ,in, is the input vector, is the center vector, for and The Euclidean distance between is a monotonic function, when keep away When the function value is very small, exp is an exponential function with the natural constant e as the base. The hyperparameters used in this embodiment are the regularization parameter C, the kernel function coefficient, etc., which are used to control the complexity of the model and prevent overfitting. A range is defined for the regularization parameter C and the kernel function coefficient respectively, and multiple candidate values ​​are randomly selected within this range to participate in the training of the subsequent model.

[0097] Determine the objective function: ,in, It is a regularization term, which makes the weight vector as small as possible, thereby controlling the complexity of the model. A small weight vector means that the model is less dependent on the training data, which may reduce the risk of overfitting. is the loss term, which measures how well the model fits the training data. C is the regularization parameter, which controls the trade-off between the regularization term and the loss term. Specifically, the larger the value of C, the more the model focuses on minimizing the training error (ie, the loss term), which may cause the model to perform well on the training data but may overfit. On the contrary, the smaller the value of C, the more the model focuses on reducing the norm of the weight vector (ie, the regularization term), which may cause the model to perform averagely on the training data but have stronger generalization ability. Therefore, by adjusting the value of C, the complexity and generalization ability of the second preset network model can be controlled. In practical applications, the value of C is usually selected through methods such as cross-validation to find a model that performs well on both the training set and the test set. w is the weight vector (the coefficient that defines the decision boundary in the feature space), b is the bias term of the decision boundary, is a slack variable (used to handle samples that do not satisfy the hard interval constraint, that is, allowing some samples to be inside the interval or on the wrong side), n is the number of training samples, and C is the regularization parameter.

[0098] After the kernel function, hyperparameters and objective function of the second preset network model are determined, the training samples are used to train it to obtain the trained second preset network model; the test samples are input into the trained second preset network model for performance evaluation, and the hyperparameters (i.e., regularization parameters C) and kernel function coefficients that make the trained second preset network model have the best performance (or minimize the objective function) are selected. After these parameters or coefficients are determined, the optimal hyperplane and defect recognition model can be determined.

[0099] In an optional embodiment of the present invention, step 10235 includes:

[0100] Step 102351, inputting the fused defect image into the input layer of the defect recognition model for processing to obtain a first output result;

[0101] Specifically, the input layer of the defect recognition model receives the fused defect image and converts the fused defect image into a feature vector, which contains key information for classification.

[0102] Step 102352, inputting the first output result into the mapping layer of the defect recognition model for processing to obtain a second output result;

[0103] Specifically, since the defects on the battery surface may appear in various complex shapes and textures, these features are often not linearly separable. Therefore, a mapping layer is needed to map the input low-dimensional nonlinear features into a high-dimensional linearly separable space, so that the defect recognition model can effectively perform defect recognition.

[0104] Step 102353, inputting the second output result into the separation layer of the defect recognition model for processing to obtain a third output result;

[0105] Specifically, in order to improve the accuracy of the final defect recognition, it is also necessary to use the separation layer of the defect recognition model to separate the normal area and the defect area in the second output result.

[0106] Step 102354, input the third output result into the output layer of the defect recognition model for processing to obtain a surface detection result.

[0107] Specifically, the output layer of the defect recognition model gives the surface detection result of the battery according to the third output result. The surface detection result includes information such as the location and type of the defect.

[0108] In an optional embodiment of the present invention, the second data to be detected is an ultrasonic signal of the lithium battery and an internal image of the lithium battery collected by an air-coupled ultrasonic detection system.

[0109] Specifically, when the air-coupled ultrasonic detection system collects the ultrasonic signal of the lithium battery, it is necessary to adjust the probe position and sampling parameters first. Here, the air-coupled transmitting probe and the receiving probe are placed vertically on the upper and lower sides of the lithium battery, and the central axis is kept coincident. The probe position is adjusted to maximize the amplitude of the received ultrasonic signal, and the relative position of the probe is fixed. The sampling parameters include sampling frequency, sampling length, gain, and various motion parameter indicators, etc. The settings of these parameters should be adjusted according to different battery specifications and detection requirements. After that, the ultrasonic signal of the lithium battery is collected, and the required excitation signal is output through the RF power amplifier to excite the high-sensitivity air-transmitting probe, convert the electrical signal into an ultrasonic signal and emit it into the air. After the ultrasonic signal penetrates the lithium battery and reaches the air, it is received by the air-coupled receiving probe and the acoustic signal is converted into an electrical signal; the preamplifier filters and amplifies the received electrical signal to remove noise and interference signals and improve the signal quality; the ultrasonic receiver converts the amplified electrical signal into a digital signal and stores it as the ultrasonic signal of the lithium battery.

[0110] The motor motion controller and the scanning frame are used to scan the lithium battery based on preset parameters to obtain the internal image of the lithium battery. The preset parameters may include scanning direction and step, scanning range, sampling frequency and length, gain and ultrasonic frequency.

[0111] In an optional embodiment of the present invention, step 103 includes:

[0112] Step 1031, extracting characteristic parameters of the ultrasonic signal of the lithium battery to obtain characteristic parameters;

[0113] Specifically, the ultrasonic signal of the lithium battery contains characteristic parameters such as the amplitude and phase of the ultrasonic waveform signal. These parameters are extracted and processed for subsequent internal detection of the lithium battery.

[0114] Step 1032: Perform an internal inspection on the lithium battery according to the characteristic parameters and the internal image of the lithium battery to obtain an internal inspection result.

[0115] Specifically, the presence of defects (such as bubbles, stratification, etc.) inside the lithium battery can be determined based on the change pattern of characteristic parameters such as the amplitude and phase of the ultrasonic waveform signal; the internal image of the lithium battery can intuitively display the distribution of defects inside the lithium battery, and the type, size and location of the defects can be determined based on the color, brightness and other characteristics of the scanned image.

[0116] Specifically, if there are defects inside the lithium battery (such as cracks, pores, stratification, etc.), the ultrasonic wave will be reflected and scattered at the defect during the propagation process, resulting in a decrease in the amplitude of the received ultrasonic signal. By comparing the ultrasonic signal amplitudes of normal lithium batteries and the lithium battery to be tested, it can be determined whether there is an attenuation abnormality. When the ultrasonic wave encounters a gaseous substance (such as the gas generated inside the lithium battery), the ultrasonic wave is almost completely reflected due to the large difference in acoustic impedance between the gaseous substance and the solid substance, resulting in a significant increase in the amplitude of the reflected signal. When the ultrasonic wave propagates inside the lithium battery, if it encounters different media or structural changes (such as stratification, cracks, etc.), the phase of the ultrasonic signal will shift. Here, a preset standard lithium battery ultrasonic signal amplitude can be set for comparison with the amplitude of the ultrasonic waveform signal in the characteristic parameters, and a preset standard lithium battery ultrasonic signal phase can be set for comparison with the phase of the ultrasonic waveform signal in the characteristic parameters to determine whether there are defects inside the lithium battery.

[0117] Specifically, lithium batteries usually have a relatively uniform color distribution inside. If there are areas in the internal image of the battery that are significantly different in color from the surrounding areas (such as dark spots, bright spots, etc.), it may be an indication of defects. For example, bubbles may appear as bright spots, while impurities or cracks may appear as dark spots; observing the gradual change of color in the image can reveal subtle changes in the internal structure of the battery. Sudden changes in color gradients may indicate defects such as stratification and cracks. An area in the internal image of the battery with too high or too low brightness compared to the surrounding area may be a defect. For example, corrosion or holes may cause a decrease in brightness, while internal short circuits or gas accumulation may cause an increase in brightness. Analyzing the distribution of brightness in the internal image of the battery can help identify large-area defects or areas where defects are concentrated. Based on the characteristics of color and brightness changes and their correspondence with the internal structure of the battery, the type of defect can be preliminarily determined. For example, areas with obvious color changes and irregular shapes may be cracks or impurities; and areas with abnormal brightness and circular or elliptical shapes may be bubbles or holes. By measuring the number of pixels or area proportion of the defect area in the image, the actual size of the defect can be estimated. This can be done with the help of measurement tools in image processing software. According to the position coordinates of the defect area in the image, the specific location of the defect inside the lithium battery can be determined.

[0118] In an optional embodiment of the present invention, the third data to be detected is the voltage data of the lithium battery collected by the voltage collection device and the battery capacity data of the lithium battery collected by the battery capacity detection device. The preset detection threshold includes a preset voltage detection threshold and a preset battery capacity detection threshold. The preset voltage detection threshold and the preset battery capacity detection threshold can be set according to the production requirements of the lithium battery or the user requirements.

[0119] Specifically, the voltage data of the lithium battery can be collected through a voltmeter or a multimeter; the battery capacity detection equipment obtains the battery capacity data by injecting an external voltage into the lithium battery and measuring the change in the reaction rate of the lithium battery.

[0120] In an optional embodiment of the present invention, step 104 includes:

[0121] Step 1041, performing a performance test on the lithium battery according to the voltage data of the lithium battery and a preset voltage test threshold to obtain a voltage test result;

[0122] Specifically, the preset voltage detection threshold may be 0 to 10 V. If the voltage data of the lithium battery is within the preset voltage detection threshold, the lithium battery may be short-circuited. The voltage detection result may include the voltage data of the lithium battery and the preset voltage detection threshold, and the result of the comparison between the two.

[0123] Step 1042, performing a performance test on the lithium battery according to the battery capacity data of the lithium battery and a preset battery capacity test threshold to obtain a battery capacity test result;

[0124] Specifically, the preset battery capacity detection threshold can be set according to the model of the lithium battery or user needs. If the battery capacity data of the lithium battery is within the preset battery capacity detection threshold, it means that the performance of the lithium battery is good and meets the expected requirements. The battery capacity detection result may include the battery capacity data of the lithium battery and the preset battery capacity detection threshold, as well as the result of the comparison between the two.

[0125] Step 1043, obtaining a performance test result according to the voltage test result and the battery capacity test result.

[0126] Specifically, the performance test results may include voltage test results and battery capacity test results, and may also include a judgment on the performance of the lithium battery (such as good performance or poor performance) based on the comparison results of the lithium battery voltage data and a preset voltage detection threshold, and the comparison results of the lithium battery capacity data and a preset battery capacity detection threshold.

[0127] In an optional embodiment of the present invention, step 105 includes:

[0128] Step 1051, obtaining a preset automatic detection result template; the preset automatic detection result template is determined according to the user's detection requirements;

[0129] Specifically, according to actual conditions, a specific preset automatic detection result template can be set, which may include data contents such as surface detection results, internal detection results and performance detection results, wherein the surface detection results include the first image data to be detected and the location, type and other information of the detected defects; the internal detection results include the second data to be detected and the type, size and location of the detected defects; the performance detection results include the third data to be detected, the voltage detection results and the battery capacity detection results, and may also include the comparison results of the voltage data of the lithium battery and the preset voltage detection threshold, the comparison results of the battery capacity data of the lithium battery and the preset battery capacity detection threshold, as well as the judgment of the performance of the lithium battery (such as good performance or poor performance), and may also include corresponding solutions, etc.

[0130] If the user only needs to inspect the surface and interior of the lithium battery, the corresponding preset automatic inspection result template may include: surface inspection result is (1), internal inspection result is (2), and defect is (3). The surface inspection result and internal inspection result of the lithium battery are then used to automatically fill in (1), (2), and (3) in the template, and the inspection result or report can be quickly generated. It should be noted that (3) is a summary of the overall defects, that is, all defects in the surface inspection results and internal inspection results can be listed in the (3) position, so that the user can intuitively see all defects.

[0131] Step 1052, extracting features from the surface detection result, the internal detection result and the performance detection result according to the preset automatic detection result template to obtain a feature set;

[0132] Specifically, according to the actual situation, the content that needs to be filled in the template is extracted from the surface inspection results, internal inspection results and performance inspection results to form a feature set, so that the inspection results or reports can be automatically and quickly generated later. For example, if a preset automatic inspection result template is: the surface inspection result is (1), the internal inspection result is (2), and the defect is (3), then the detected defect is extracted from the surface inspection result and the internal inspection result, and the surface inspection result, the internal inspection result and the defect are used as features to form a feature set for the subsequent filling of (1), (2) and (3) in the automatic inspection result template. Here, each feature in the feature set can be identified, such as (1) surface inspection result, (2) internal inspection result, (3) defect, and then (1) (2) (3) is correspondingly filled into the automatic inspection result template according to the identification.

[0133] Step 1053, filling the preset automatic detection result template according to the feature set to obtain the automatic detection result of the lithium battery.

[0134] Specifically, according to the serial number identification of each feature in the feature set, each feature is filled into the position of the corresponding serial number in the template, and the preset automatic detection result template is automatically filled to obtain the automatic detection result of the lithium battery.

[0135] A specific embodiment of the battery detection method proposed in the embodiment of the present invention is:

[0136] Step 111, obtaining the data to be tested of the lithium battery.

[0137] The surface image of the lithium battery is collected by using a camera for surface inspection of the lithium battery.

[0138] Use the air-coupled ultrasonic detection system to collect the ultrasonic signal and internal image of the lithium battery. Before collecting, check that all the hardware equipment of the air-coupled ultrasonic detection system (such as RF power amplifier, preamplifier, air-coupled probe, data acquisition card, computer, etc.) is in good working condition, check and calibrate the test instruments to ensure the accuracy and reliability of the test results. Select a suitable test environment to ensure that the temperature, humidity and other conditions meet the test requirements to reduce the impact of external factors on the test results. After that, adjust the probe position and sampling parameters in the air-coupled ultrasonic detection system. Here, the air-coupled transmitting probe and the receiving probe are placed vertically on the upper and lower sides of the lithium battery, and the central axis is kept coincident. Adjust the probe position to maximize the amplitude of the received ultrasonic signal and fix the relative position of the probe unchanged. The sampling parameters include sampling frequency, sampling length, gain, and various motion parameter indicators. The settings of these parameters should be adjusted according to different battery specifications and detection requirements. After that, the ultrasonic signal of the lithium battery is collected, and the required excitation signal is output through the RF power amplifier to excite the high-sensitivity air-transmitting probe, convert the electrical signal into an ultrasonic signal and transmit it into the air. The ultrasonic signal penetrates the lithium battery and reaches the air, where it is received by the air-coupled receiving probe and converts the acoustic signal into an electrical signal. The preamplifier filters and amplifies the received electrical signal to remove noise and interference signals and improve signal quality. The ultrasonic receiver converts the amplified electrical signal into a digital signal and stores it as the ultrasonic signal of the lithium battery. The lithium battery is scanned according to the preset parameters by the motor motion controller and the scanning frame to obtain the internal image of the lithium battery. The preset parameters may include scanning direction and step, scanning range, sampling frequency and length, gain, and ultrasonic frequency.

[0139] In this embodiment, the scanning direction generally includes the X direction and the Y direction, which determines the moving trajectory of the probe on the surface of the lithium battery; the step distance is the distance of each movement, such as 0.3mm, which affects the resolution and accuracy of the scan. A smaller step distance can provide a more detailed detection image, but it will also increase the detection time and data volume. The scanning range defines the entire area that the probe needs to scan, such as a range of 268mm×368mm, to ensure that all key areas of the lithium battery are included for comprehensive detection of internal defects. The sampling frequency determines the number of ultrasonic signal samples collected per second, such as a sampling frequency of 10MHz, which can provide a higher time resolution and help capture subtle ultrasonic changes. The sampling length is the number of samples obtained per sampling, such as 3584 points collected per group, which affects the integrity of the signal and the detection accuracy. Gain is the process of amplifying the ultrasonic signal for clearer detection and analysis. In air-coupled ultrasonic detection, an appropriate gain setting (such as 33dB) can ensure that the signal is neither too strong nor too weak, thereby improving the accuracy of detection. The frequency selection of ultrasound has a great influence on the detection effect. Air-coupled ultrasound generally uses probes with a frequency of tens of kHz to hundreds of kHz. Because high-frequency sound waves will be absorbed after traveling a very short distance in the air, it is necessary to select a suitable ultrasonic frequency based on the characteristics of the lithium battery and the detection requirements.

[0140] The voltage data of the lithium battery is collected through a voltmeter or a multimeter; the battery capacity detection equipment obtains the battery capacity data by injecting an external voltage into the lithium battery and measuring the change in the reaction rate of the lithium battery.

[0141] Step 112: Lithium battery surface inspection.

[0142] First, the surface image of the lithium battery is preprocessed by correction, and then the area to be processed containing the lithium battery is extracted from it. The area to be processed is input into the pre-trained target area recognition model for target area recognition, and the target area including the names of various parts of the lithium battery (such as pole pieces, lithium battery body, etc.) and the corresponding positions is obtained. After at least one defect sub-image is determined according to the target area, the defect sub-image is fused and filtered, and then the surface of the lithium battery is inspected using the trained defect recognition model to obtain the lithium battery surface inspection result containing information such as the location and type of the defect.

[0143] Step 113: internal inspection of the lithium battery.

[0144] The characteristic parameters such as the amplitude and phase of the ultrasonic waveform signal are extracted from the ultrasonic signal of the lithium battery. Then, based on the change pattern of these parameters, it is judged whether there are defects (such as bubbles, stratification, etc.) inside the lithium battery. According to the color, brightness and other characteristics of the scanned image of the lithium battery, the type, size and location of the defects are judged, and the internal detection results of the lithium battery including whether there are defects inside, as well as the type, size, location and other information of the defects are obtained.

[0145] Step 114, lithium battery performance testing.

[0146] The voltage data of the lithium battery is compared with the preset voltage detection threshold, and the battery capacity data of the lithium battery is compared with the preset battery capacity detection threshold to obtain the lithium battery performance test result of whether the battery voltage and battery capacity meet the expected requirements.

[0147] Step 115, outputting the lithium battery automatic detection result.

[0148] According to the actual situation, a preset automatic test result template can be set or selected, and the template can include the data to be tested of the lithium battery, as well as all the data content in the surface test results, internal test results and performance test results, and can also include corresponding solutions, etc. The information required by the preset automatic test result template is extracted from the surface test results, internal test results and performance test results, and the preset automatic test result template is filled to obtain the automatic test result of the lithium battery.

[0149] The battery detection method proposed in the embodiment of the present invention performs surface, internal and performance detection on the lithium battery by acquiring the data of the lithium battery to be detected, and can comprehensively and automatically make an automatic detection result on whether the lithium battery has defects and whether it meets the expected production requirements, which is beneficial to improving the production efficiency of lithium batteries, reducing manual workload, and improving the automation level of lithium battery detection.

[0150] like Figure 4 As shown, an embodiment of the present invention provides a battery detection device 200, including:

[0151] An acquisition module 201 is used to acquire first image data to be detected, second data to be detected and third data to be detected of a lithium battery;

[0152] A first detection module 202, configured to perform surface detection on the lithium battery according to the first to-be-detected image data and a pre-trained target region recognition model to obtain a surface detection result;

[0153] A second detection module 203, configured to perform an internal detection on the lithium battery according to the second to-be-detected data to obtain an internal detection result;

[0154] A third detection module 204, configured to perform a performance test on the lithium battery according to the third to-be-detected data and a preset detection threshold, and obtain a performance test result;

[0155] The processing module 205 is used to fill in a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain the automatic detection result of the lithium battery.

[0156] Optionally, the first image data to be detected is a surface image of a lithium battery collected by a camera device, and the target area recognition model is obtained by training a first preset network model using a battery feature image and its corresponding label; performing surface detection on the lithium battery according to the first image data to be detected and the pre-trained target area recognition model to obtain a surface detection result includes:

[0157] Correcting the surface image of the lithium battery to obtain a corrected surface image;

[0158] Performing region extraction on the corrected surface image to obtain a region to be processed;

[0159] The lithium battery is subjected to surface detection according to the area to be processed and a pre-trained target area recognition model to obtain a surface detection result.

[0160] Optionally, performing surface detection on the lithium battery according to the area to be processed and a pre-trained target area recognition model to obtain a surface detection result includes:

[0161] Inputting the area to be processed into a pre-trained target area recognition model to perform target area recognition to obtain a target area;

[0162] Performing grayscale adjustment on the target area to obtain an adjusted target area;

[0163] Determining a defect image according to a preset threshold and the adjusted target area;

[0164] Performing fusion processing on the defect image to obtain a fused defect image;

[0165] According to the fused defect image and the trained defect recognition model, the lithium battery is subjected to surface inspection to obtain a surface inspection result; the defect recognition model is obtained by training a second preset network model using the surface defect image of the lithium battery.

[0166] Optionally, performing surface inspection on the lithium battery according to the fused defect image and the trained defect recognition model to obtain a surface inspection result includes:

[0167] Inputting the fused defect image into the input layer of the defect recognition model for processing to obtain a first output result;

[0168] Inputting the first output result into a mapping layer of a defect recognition model for processing to obtain a second output result;

[0169] Inputting the second output result into the separation layer of the defect recognition model for processing to obtain a third output result;

[0170] The third output result is input into the output layer of the defect recognition model for processing to obtain a surface detection result.

[0171] Optionally, the second data to be detected is an ultrasonic signal of a lithium battery and an internal image of the lithium battery collected by an air-coupled ultrasonic detection system; performing internal detection on the lithium battery according to the second data to be detected to obtain an internal detection result includes:

[0172] Extracting characteristic parameters of the ultrasonic signal of the lithium battery to obtain characteristic parameters;

[0173] An internal inspection of the lithium battery is performed according to the characteristic parameters and the internal image of the lithium battery to obtain an internal inspection result.

[0174] Optionally, the third data to be detected is voltage data of the lithium battery collected by a voltage collection device and battery capacity data of the lithium battery collected by a battery capacity detection device, and the preset detection threshold includes a preset voltage detection threshold and a preset battery capacity detection threshold; and the lithium battery is subjected to performance detection according to the third data to be detected and the preset detection threshold to obtain a performance detection result, including:

[0175] Performing a performance test on the lithium battery according to the voltage data of the lithium battery and a preset voltage detection threshold to obtain a voltage detection result;

[0176] Performing a performance test on the lithium battery according to the battery capacity data of the lithium battery and a preset battery capacity test threshold to obtain a battery capacity test result;

[0177] A performance test result is obtained according to the voltage test result and the battery capacity test result.

[0178] Optionally, filling a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain the automatic detection result of the lithium battery includes:

[0179] Get the preset automatic detection result template;

[0180] Extracting features from the surface detection result, the internal detection result, and the performance detection result according to the preset automatic detection result template to obtain a feature set;

[0181] The preset automatic detection result template is filled according to the feature set to obtain the automatic detection result of the lithium battery.

[0182] The battery detection device of the embodiment of the present invention realizes surface detection, internal detection and performance detection of the lithium battery by acquiring first image data to be detected, second data to be detected and third data to be detected of the lithium battery, and automatically processes these data by using a pre-trained target area recognition model and a preset detection threshold, etc., and obtains automatic detection results of the lithium battery based on the obtained surface detection results, internal detection results and performance detection results, thereby reducing manual participation, which can not only realize automatic detection of the lithium battery, but also improve the comprehensiveness and accuracy of lithium battery detection.

[0183] It should be noted that the device is a device corresponding to the above method, and all implementations in the above method embodiment are applicable to the embodiment of the device and can achieve the same technical effect. This embodiment will not be repeated.

[0184] The embodiment of the present invention further provides a computing device, comprising: a processor, and a memory storing a computer program, wherein when the computer program is executed by the processor, the method described in any one of the above embodiments is executed. All implementations in the above method embodiments are applicable to the embodiments of the device, and can achieve the same technical effects. They will not be described in detail in this embodiment.

[0185] The embodiment of the present invention further provides a computer-readable storage medium on which instructions are stored. When the instructions are executed on a computer, the computer executes the method as described in any one of the above embodiments. All implementations in the above method embodiments are applicable to the embodiments of the device and can achieve the same technical effects. They will not be described in detail in this embodiment.

[0186] It should be noted that, in the above-mentioned embodiments, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "include one..." do not exclude the existence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the implementation of the above-mentioned embodiments is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0187] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A battery detection method, characterized in that: include: Acquire first image data to be detected, second data to be detected, and third data to be detected of the lithium battery; Performing surface detection on the lithium battery according to the first to-be-detected image data and a pre-trained target area recognition model to obtain a surface detection result; Performing an internal test on the lithium battery according to the second data to be tested to obtain an internal test result; Performing a performance test on the lithium battery according to the third data to be tested and a preset test threshold to obtain a performance test result; Filling a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain an automatic detection result of the lithium battery; The first image data to be detected is a surface image of a lithium battery collected by a camera device, and the target area recognition model is obtained by training a first preset network model using a battery feature image and its corresponding label; the surface of the lithium battery is detected according to the first image data to be detected and the pre-trained target area recognition model to obtain a surface detection result, including: Correcting the surface image of the lithium battery to obtain a corrected surface image; the correction method includes at least one of geometric correction, radiation correction and color correction; Performing region extraction on the corrected surface image to obtain a region to be processed; wherein, performing grayscale processing on the corrected surface image and converting it into a binary image, searching for contours in the binary image according to the edges in the binary image, and screening according to the shape, size, and direction attributes of the contours; determining a target region according to the screened contours; and cutting out the target region from the corrected surface image according to the coordinates of the target region to obtain a region to be processed; Performing surface detection on the lithium battery according to the area to be processed and a pre-trained target area recognition model to obtain a surface detection result; The target area recognition model is obtained by training the first preset network model using the battery feature image and its corresponding label as a data set; the training process includes: Collect images containing the appearance of lithium batteries and corresponding labels as training samples; Assign random values ​​to all weight vectors and biases in the first preset network model and define the loss function , where L is the loss function, is the output result of the model, and y is the real result; Inputting the training sample into the first preset network model for training until the loss function reaches a preset minimum value, stopping the training, and obtaining a target weight vector and a target bias; Obtaining a target region recognition model according to the target weight vector and the target bias; The surface detection of the lithium battery is performed according to the area to be processed and the pre-trained target area recognition model to obtain the surface detection result, including: Inputting the area to be processed into a pre-trained target area recognition model to perform target area recognition to obtain a target area; Performing grayscale adjustment on the target area to obtain an adjusted target area; Determining a defect image according to a preset threshold and the adjusted target area; Performing fusion processing on the defect image to obtain a fused defect image; According to the fused defect image and the trained defect recognition model, the lithium battery is subjected to surface inspection to obtain a surface inspection result; the defect recognition model is obtained by training a second preset network model using the surface defect image of the lithium battery; the training process includes: Collecting and preprocessing the surface defect image of the lithium battery to obtain the preprocessed surface defect image of the battery; According to the preprocessed battery surface defect image, a training sample and a test sample are obtained; the training sample is used to perform a kernel function The objective function is The second preset network model is trained to obtain the trained second preset network model; wherein, is the input vector, is the center vector, for and The Euclidean distance between is a monotonic function, when keep away When the function value is very small, exp is an exponential function with the natural constant e as the base; is the regularization term, C is the regularization parameter, is the loss term, w is the weight vector, b is the bias term of the decision boundary, is the slack variable, n is the number of training samples; Inputting the test sample into the trained second preset network model for performance evaluation, and determining a defect recognition model according to a regularization parameter and a kernel function coefficient that enable the trained second preset network model to have a minimum objective function; According to the fused defect image and the trained defect recognition model, the surface of the lithium battery is inspected to obtain the surface inspection result, including: Inputting the fused defect image into the input layer of the defect recognition model for processing to obtain a first output result; Inputting the first output result into a mapping layer of a defect recognition model for processing to obtain a second output result; Inputting the second output result into the separation layer of the defect recognition model for processing to obtain a third output result; The third output result is input into the output layer of the defect recognition model for processing to obtain a surface detection result.

2. The battery detection method according to claim 1, characterized in that: The second data to be detected is an ultrasonic signal of the lithium battery and an internal image of the lithium battery collected by an air-coupled ultrasonic detection system; Performing an internal test on the lithium battery according to the second to-be-tested data to obtain an internal test result, including: Extracting characteristic parameters of the ultrasonic signal of the lithium battery to obtain characteristic parameters; An internal inspection of the lithium battery is performed according to the characteristic parameters and the internal image of the lithium battery to obtain an internal inspection result.

3. The battery detection method according to claim 1, characterized in that: The third data to be detected is the voltage data of the lithium battery collected by the voltage collection device and the battery capacity data of the lithium battery collected by the battery capacity detection device, and the preset detection threshold includes a preset voltage detection threshold and a preset battery capacity detection threshold; The lithium battery is subjected to a performance test according to the third to-be-tested data and a preset test threshold value to obtain a performance test result, including: Performing a performance test on the lithium battery according to the voltage data of the lithium battery and a preset voltage detection threshold to obtain a voltage detection result; Performing a performance test on the lithium battery according to the battery capacity data of the lithium battery and a preset battery capacity test threshold to obtain a battery capacity test result; A performance test result is obtained according to the voltage test result and the battery capacity test result.

4. The battery detection method according to claim 1, characterized in that: Filling a preset automatic test result template according to the surface test result, the internal test result and the performance test result to obtain the automatic test result of the lithium battery, including: Get the preset automatic detection result template; Extracting features from the surface detection result, the internal detection result, and the performance detection result according to the preset automatic detection result template to obtain a feature set; The preset automatic detection result template is filled according to the feature set to obtain the automatic detection result of the lithium battery.

5. A battery detection device, characterized in that: include: An acquisition module, used to acquire first image data to be detected, second data to be detected and third data to be detected of the lithium battery; A first detection module, used to perform surface detection on the lithium battery according to the first to-be-detected image data and a pre-trained target area recognition model to obtain a surface detection result; A second detection module, used for performing an internal detection on the lithium battery according to the second data to be detected, and obtaining an internal detection result; A third detection module, used to perform a performance test on the lithium battery according to the third to-be-detected data and a preset detection threshold, to obtain a performance test result; A processing module, used to fill a preset automatic detection result template according to the surface detection result, the internal detection result and the performance detection result to obtain an automatic detection result of the lithium battery; The first image data to be detected is a surface image of a lithium battery collected by a camera device, and the target area recognition model is obtained by training a first preset network model using a battery feature image and its corresponding label; the surface of the lithium battery is detected according to the first image data to be detected and the pre-trained target area recognition model to obtain a surface detection result, including: Correcting the surface image of the lithium battery to obtain a corrected surface image; the correction method includes at least one of geometric correction, radiation correction and color correction; Performing region extraction on the corrected surface image to obtain a region to be processed; wherein, performing grayscale processing on the corrected surface image and converting it into a binary image, searching for contours in the binary image according to the edges in the binary image, and screening according to the shape, size, and direction attributes of the contours; determining a target region according to the screened contours; and cutting out the target region from the corrected surface image according to the coordinates of the target region to obtain a region to be processed; Performing surface detection on the lithium battery according to the area to be processed and a pre-trained target area recognition model to obtain a surface detection result; The target area recognition model is obtained by training the first preset network model using the battery feature image and its corresponding label as a data set; the training process includes: Collect images containing the appearance of lithium batteries and corresponding labels as training samples; Assign random values ​​to all weight vectors and biases in the first preset network model and define the loss function , where L is the loss function, is the output result of the model, and y is the real result; Inputting the training sample into the first preset network model for training until the loss function reaches a preset minimum value, stopping the training, and obtaining a target weight vector and a target bias; Obtaining a target region recognition model according to the target weight vector and the target bias; The surface detection of the lithium battery is performed according to the area to be processed and the pre-trained target area recognition model to obtain the surface detection result, including: Inputting the area to be processed into a pre-trained target area recognition model to perform target area recognition to obtain a target area; Performing grayscale adjustment on the target area to obtain an adjusted target area; Determining a defect image according to a preset threshold and the adjusted target area; Performing fusion processing on the defect image to obtain a fused defect image; According to the fused defect image and the trained defect recognition model, the lithium battery is subjected to surface inspection to obtain a surface inspection result; the defect recognition model is obtained by training a second preset network model using the surface defect image of the lithium battery; the training process includes: Collecting and preprocessing the surface defect image of the lithium battery to obtain the preprocessed surface defect image of the battery; According to the preprocessed battery surface defect image, a training sample and a test sample are obtained; the training sample is used to perform a kernel function The objective function is The second preset network model is trained to obtain the trained second preset network model; wherein, is the input vector, is the center vector, for and The Euclidean distance between is a monotonic function, when keep away When the function value is very small, exp is an exponential function with the natural constant e as the base; is the regularization term, C is the regularization parameter, is the loss term, w is the weight vector, b is the bias term of the decision boundary, is the slack variable, n is the number of training samples; Inputting the test sample into the trained second preset network model for performance evaluation, and determining a defect recognition model according to a regularization parameter and a kernel function coefficient that enable the trained second preset network model to have a minimum objective function; According to the fused defect image and the trained defect recognition model, the surface of the lithium battery is inspected to obtain the surface inspection result, including: Inputting the fused defect image into the input layer of the defect recognition model for processing to obtain a first output result; Inputting the first output result into a mapping layer of a defect recognition model for processing to obtain a second output result; Inputting the second output result into the separation layer of the defect recognition model for processing to obtain a third output result; The third output result is input into the output layer of the defect recognition model for processing to obtain a surface detection result.

6. A computing device, characterized in that include: A processor and a memory storing a computer program, wherein when the computer program is executed by the processor, the battery detection method according to any one of claims 1 to 4 is executed.

7. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer is caused to execute the battery detection method according to any one of claims 1 to 4.

Images