Glue testing methods, devices, equipment, media and products

Abstract

The present application discloses a method, device, equipment, medium and product for detecting glue sticking, wherein the method comprises: obtaining a first battery cell image of a battery cell to be detected; using a trained image segmentation model to segment the glue sticking area in the first battery cell image to obtain a first mask image; the trained image segmentation model is trained by the following model training method: obtaining a second battery cell image; using the image segmentation model to be trained, segmenting the glue sticking area in the second battery cell image to obtain a second mask image and a predicted film type; when the second mask image does not meet the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type; using the second battery cell image and the target film type, training the image segmentation model to be trained to obtain a trained image segmentation model; based on the first mask image, performing glue sticking detection on the first battery cell image to obtain a glue sticking detection result.

Description

Technical Field

[0001] The present disclosure relates to but is not limited to the field of machine vision detection technology, and in particular to a method, device, equipment, medium and product for detecting adhesive bonding. Background Art

[0002] In order to detect whether the glue area in the positive and negative electrodes of lithium battery welding covers the weld marks, it is necessary to segment the glue area to detect whether the glue coverage is abnormal. In related technologies, the blue channel can be used to detect the glue area. This method often requires modifying the threshold of the blue channel due to changes in factors such as the lighting conditions and equipment status of the production line. To address this problem, network segmentation (such as Unet network segmentation) is used to detect the glue area, but this method often has insufficient segmentation accuracy and is prone to overkill. Summary of the invention

[0003] In view of this, the embodiments of the present disclosure at least provide a method, device, equipment, medium and product for detecting adhesive sticking.

[0004] The technical solution of the embodiment of the present disclosure is implemented as follows:

[0005] The present disclosure provides a method for detecting adhesive sticking, the method comprising:

[0006] Acquire a first battery cell image of a battery cell to be detected;

[0007] Using the trained image segmentation model, the glue area in the first battery cell image is segmented to obtain a first segmentation result, and the first segmentation result includes a first mask image corresponding to the glue area; the trained image segmentation model is trained using the following model training method: obtaining a second battery cell image; using the image segmentation model to be trained, the glue area in the second battery cell image is segmented to obtain a second segmentation result, and the second segmentation result includes a second mask image corresponding to the glue area and a predicted film type; when the second mask image does not meet the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type; using the second battery cell image and the target film type, the image segmentation model to be trained is trained to obtain the trained image segmentation model;

[0008] Based on the first mask image, a glue sticking detection is performed on the first battery cell image to obtain a glue sticking detection result.

[0009] In the disclosed embodiment, the image segmentation model to be trained is trained by the following model training method: the image segmentation model to be trained is used to segment the glue area in the second battery cell image to obtain the second mask image corresponding to the glue area and the predicted film type; when it is determined that the second mask image does not meet the accuracy condition, the target film type different from the predicted film type is determined, and the second battery cell image and the target film type are used to train the image segmentation model to be trained to obtain the trained image segmentation model. The trained image segmentation model is used to segment the first battery cell image to be detected to obtain the first mask image, and the first battery cell image is detected according to the first mask image to obtain the glue detection result. In this way, on the one hand, the image segmentation model to be trained according to different film types can improve the efficiency of model training, and make the trained image segmentation model compatible with different types of film types; on the other hand, when the second mask image does not meet the accuracy condition, the image segmentation model is retrained by changing the film type, so that the image segmentation model obtained after training is more accurate, thereby improving the accuracy of glue detection.

[0010] In some embodiments, after acquiring the first cell image of the cell to be detected, the method further includes: resetting the size of the first cell image to the first size, and performing pixel value padding on the first cell image after the size is reset, and expanding the size of the first cell image to the second size; wherein the second size is larger than the first size; the second size is the size required by the trained image segmentation model to segment the first cell image; using the trained image segmentation model to segment the glue area in the first cell image to obtain the first segmentation result includes: using the trained image segmentation model to segment the glue area in the first cell image after the size is expanded, and obtaining the first segmentation result. In this way, the size of the first cell image is first reset to the first size, and then the first size is expanded to the second size by filling the pixel values. This processing method can maintain the ratio of the height and width of the first cell image and prevent the loss of image accuracy.

[0011] In some embodiments, the use of the trained image segmentation model to segment the glue area in the first battery cell image to obtain a first segmentation result includes: using the trained image segmentation model to perform target detection on the glue area in the first battery cell image to obtain a target detection result, wherein the target detection result includes a target detection frame corresponding to the glue area; based on the target detection frame, the glue area in the first battery cell image is segmented to determine a first mask image corresponding to the glue area. In this way, the glue area is segmented according to the target detection frame obtained by target detection to obtain a first mask image; the target position of the glue area and the shape of the glue area can be obtained according to the first mask image.

[0012] In some embodiments, the use of the trained image segmentation model to perform target detection on the glue area in the first battery cell image to obtain a target detection result includes: using the trained image segmentation model to perform target detection on the glue area in the first battery cell image to obtain at least one candidate detection frame corresponding to the glue area; based on the at least one candidate detection frame, determining a target detection frame corresponding to the glue area. In this way, a target detection frame is determined from multiple candidate detection frames obtained by target detection, and the target detection frame can accurately locate the target position of the glue area, making the detection result of the glue detection more accurate.

[0013] In some embodiments, determining the target detection frame corresponding to the glued area based on the at least one candidate detection frame includes: determining the confidence corresponding to each of the candidate detection frames; based on the confidence corresponding to each of the candidate detection frames, selecting at least one first target detection frame from each of the candidate detection frames; performing non-maximum suppression on each of the first target detection frames to determine the target detection frame. In this way, the first target detection frame with a greater possibility is selected according to the confidence corresponding to each of the candidate detection frames; performing non-maximum suppression on the first target detection frame to remove duplicate first target detection frames, and obtaining the final target detection frame.

[0014] In some embodiments, the candidate types include a first film type of the film pasted on the front of the battery cell and a second film type of the film pasted on the back of the battery cell; the determination of a target film type different from the predicted film type from at least one candidate type includes at least one of the following: when the predicted film type is the first film type, determining the target film type is the second film type; when the predicted film type is the second film type, determining the target film type is the first film type. In this way, by changing the film type and retraining the image segmentation model, a more accurate image segmentation model is obtained, thereby improving the accuracy of using the image segmentation model to detect the adhesive.

[0015] In some embodiments, the film pasted on the front of the battery cell includes a film of a single color, and the film pasted on the back of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color; or, the film pasted on the front of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color, and the film pasted on the surface of the battery cell includes a film of a single color. In this way, the film colors of the film pasted on the front of the battery cell and the film pasted on the back of the battery cell are different, so that the features of the film pasted on the front of the battery cell and the film pasted on the back of the battery cell are obviously different, so that the image segmentation model is trained using the second battery cell image and the target film type, which can improve the image segmentation model's ability to recognize the film pasted on the front of the battery cell and the film pasted on the back of the battery cell.

[0016] In some embodiments, the second cell image has a glue area label; the method further includes: if the second mask image does not match the glue area label, determining that the second mask image does not meet the accuracy condition. In this way, the glue area label is used as reference information, and the target position of the glue area in the second mask image is compared with the reference information to determine whether the second mask image meets the accuracy condition. This method can accurately determine whether the second mask image meets the accuracy condition.

[0017] In some embodiments, the method further comprises: obtaining a determination result in response to an accuracy determination operation on the second mask image; and determining whether the second mask image satisfies an accuracy condition based on the determination result. In this way, by performing an accuracy determination operation on the second mask image to determine whether the second mask image satisfies the accuracy condition, this method can quickly determine whether the second mask image satisfies the accuracy condition.

[0018] In some embodiments, the method further includes: obtaining the film color information of the first battery cell image; the film color information includes at least two colors or a single color; the using of the image segmentation model to segment the glue area in the first battery cell image to obtain a first segmentation result includes at least one of the following: when the film color information is at least two colors, using the image segmentation model to segment the glue area in the first battery cell image to obtain a first segmentation result; when the film color information is a single color, using the color channel corresponding to the single color to segment the glue area in the first battery cell image to obtain a first segmentation result. In this way, the glue area of ​​a single color is directly segmented using the blue channel, which is simpler and faster; the glue area of ​​at least two colors is segmented using the image segmentation model, which is compatible with glue areas of multiple colors, and the first segmentation result obtained is more accurate.

[0019] In some embodiments, the method of performing glue detection on the first battery cell image based on the first mask image to obtain a glue detection result includes: setting the glue area in the first battery cell image as a target pixel value based on the first mask image to obtain a third battery cell image; and performing glue detection on the battery cell to be detected based on the third battery cell image to obtain a glue detection result. In this way, by setting the pixel value of the glue area as the target pixel value based on the first mask image, the glue area can be better distinguished from other backgrounds, making the glue detection result more accurate.

[0020] The present disclosure provides a device for detecting adhesive sticking, the device comprising:

[0021] A first acquisition module, used to acquire a first battery cell image of a battery cell to be detected;

[0022] A first segmentation module is used to segment the glue area in the first battery cell image using a trained image segmentation model to obtain a first segmentation result, wherein the first segmentation result includes a first mask image corresponding to the glue area; the trained image segmentation model is trained using the following model training method: obtaining a second battery cell image; using the image segmentation model to be trained, segmenting the glue area in the second battery cell image to obtain a second segmentation result, wherein the second segmentation result includes a second mask image corresponding to the glue area and a predicted film type; when the second mask image does not meet the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type; using the second battery cell image and the target film type, training the image segmentation model to be trained to obtain the trained image segmentation model;

[0023] The detection module is used to perform glue sticking detection on the first battery cell image based on the first mask image to obtain a glue sticking detection result.

[0024] An embodiment of the present disclosure provides a computer device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the program, some or all of the steps in the above method are implemented.

[0025] An embodiment of the present disclosure provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, some or all of the steps in the above method are implemented.

[0026] An embodiment of the present disclosure provides a computer program product, including a computer program or instructions, which implements some or all of the steps in the above method when executed by a processor.

[0027] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used to illustrate the technical solutions of the present disclosure together with the specification.

[0029] Figure 1 A schematic diagram of the implementation process of a glue detection method provided in the embodiment of the present disclosure Figure 1 ;

[0030] Figure 2 A schematic diagram of the implementation process of a glue detection method provided in the embodiment of the present disclosure Figure 2 ;

[0031] Figure 3 A schematic diagram of the implementation flow of a method for segmenting the adhesive bonding area of ​​a battery cell provided in an embodiment of the present disclosure;

[0032] Figure 4 A glue-attached image of a battery cell to be segmented using the Unet network segmentation method provided in an embodiment of the present disclosure;

[0033] Figure 5 The glue-attached image of the battery cell to be segmented using the Yolov5-seg model provided in the embodiment of the present disclosure;

[0034] Figure 6 The segmentation result using the Unet network segmentation method provided in the embodiment of the present disclosure;

[0035] Figure 7 The segmentation result using the Yolov5-seg model provided in the embodiment of the present disclosure;

[0036] Figure 8 A schematic diagram of a film attached to the front of a battery cell provided in an embodiment of the present disclosure;

[0037] Fig. 9 A schematic diagram of a film attached to the back of a battery cell provided in an embodiment of the present disclosure;

[0038] Fig.10 The glue-attached image of the dirty battery cell to be inspected provided in the embodiment of the present disclosure;

[0039] Fig.11 The segmentation result of the adhesive image of the dirty battery cell to be inspected provided by the embodiment of the present disclosure;

[0040] Fig.12 A schematic diagram of the structure of a glue detection device provided in an embodiment of the present disclosure;

[0041] Fig.13 A hardware entity schematic diagram of a computer device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the technical solutions of the present disclosure are further elaborated in detail below in conjunction with the drawings and embodiments. The described embodiments should not be regarded as limiting the present disclosure. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0043] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0044] The terms "first / second / third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that "first / second / third" can be interchanged with a specific order or sequence where permitted so that the embodiments of the present disclosure described herein can be implemented in an order other than that illustrated or described herein.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present disclosure belongs. The terms used herein are only for the purpose of describing the present disclosure and are not intended to limit the present disclosure.

[0046] In view of this, an embodiment of the present disclosure provides a method for detecting adhesive sticking. Figure 1 A schematic diagram of the implementation process of a glue detection method provided in the embodiment of the present disclosure Figure 1 ,like Figure 1 As shown, the method includes the following steps S101 to S103:

[0047] Step S101: Acquire a first battery cell image of a battery cell to be detected;

[0048] Here, the first battery cell image is an image of a battery cell to be inspected during an actual production process.

[0049] In some embodiments, after the first battery cell image to be detected is acquired, the first battery cell image is preprocessed, that is, the size of the first battery cell image is reset according to the size of the input image required by the trained image segmentation model.

[0050] Step S102: using the trained image segmentation model, segmenting the glue area in the first battery cell image to obtain a first segmentation result, the first segmentation result includes a first mask image corresponding to the glue area; the trained image segmentation model is trained using the following model training method: obtaining a second battery cell image; using the image segmentation model to be trained, segmenting the glue area in the second battery cell image to obtain a second segmentation result, the second segmentation result includes a second mask image corresponding to the glue area and a predicted film type; when the second mask image does not meet the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type; using the second battery cell image and the target film type, training the image segmentation model to be trained to obtain the trained image segmentation model;

[0051] Here, the first segmentation result is used to detect whether the adhesive in the adhesive area of ​​the first battery cell image covers the weld mark.

[0052] In some embodiments, the trained image segmentation model may be a deep learning model, for example, the Yolov5-seg model (You Only Look Once version 5–Segmentation).

[0053] In some implementations, the glued area in the first battery cell image may be segmented using a blue channel.

[0054] In some embodiments, the adhesive in the adhesive region in the first battery cell image may be a single-color all-blue adhesive;

[0055] In some embodiments, the glue in the glue area in the first battery cell image may also include multiple colors, such as blue glue and transparent glue.

[0056] It can be understood that the first battery cell image is used as the input of the trained image segmentation model, the glue area in the first battery cell image is segmented, and the first mask image corresponding to the glue area is output. The first mask image here includes the position and shape of the glue area, which can be used to determine whether the glue in the glue area of ​​the first battery cell image completely covers the weld mark of the battery cell.

[0057] The training process of the image segmentation model to be trained may include the following steps S121 to S124:

[0058] Step S121: acquiring a second battery cell image;

[0059] Here, the second battery cell image is a sample image for training the image segmentation model to be trained.

[0060] In some embodiments, the second battery cell image is augmented by scaling, rotating, shading, etc., so as to achieve a good detection effect on the glue-attached areas of battery cells of different shapes.

[0061] In some embodiments, the sample is expanded by adding scenes such as overexposure, dirt, and spots to the second cell image.

[0062] Step S122: using the image segmentation model to be trained, segmenting the glue-sticking area in the second battery cell image to obtain a second segmentation result, wherein the second segmentation result includes a second mask image corresponding to the glue-sticking area and a predicted film type;

[0063] In some embodiments, the predicted film type may include a film attached to the front side of the battery cell and a film attached to the back side of the battery cell.

[0064] It can be understood that the second battery cell image is used as the input of the image segmentation model to be trained, the glue area in the second battery cell image is segmented, and the second mask image corresponding to the glue area and the predicted film type are output. The second mask image here includes the position and shape of the glue area, which can be used to compare with the glue area of ​​the battery cell in the second battery cell image to determine whether the second segmentation result is accurate.

[0065] Step S123: when the second mask image does not satisfy the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type;

[0066] In some embodiments, during the model training process, the second mask image and the second battery cell image may be compared to determine whether the second mask image meets the accuracy condition.

[0067] In some implementations, during the actual production process, it may be manually determined whether the second mask image meets the accuracy condition.

[0068] In some embodiments, the candidate types may include a first film type of a film attached to the front side of the battery cell and a second film type of a film attached to the back side of the battery cell.

[0069] In some embodiments, the second mask image does not meet the accuracy condition, and when the predicted film type is the first film type, the target film type is determined to be the second film type; when the predicted film type is the second film type, the target film type is determined to be the first film type.

[0070] It can be understood that if the second mask image does not meet the accuracy condition, it is necessary to select a film type different from the predicted film type corresponding to the glue area of ​​the second battery cell image as the target film type, and use the second battery cell image and the target film type as new samples to re-train the image segmentation model to be trained.

[0071] Step S124: using the second battery cell image and the target film type, training the image segmentation model to be trained to obtain the trained image segmentation model.

[0072] Step S103: Based on the first mask image, perform glue detection on the first battery cell image to obtain a glue detection result.

[0073] In the disclosed embodiment, the image segmentation model to be trained is trained by the following model training method: the image segmentation model to be trained is used to segment the glue area in the second battery cell image to obtain the second mask image corresponding to the glue area and the predicted film type; when it is determined that the second mask image does not meet the accuracy condition, the target film type different from the predicted film type is determined, and the second battery cell image and the target film type are used to train the image segmentation model to be trained to obtain the trained image segmentation model. The trained image segmentation model is used to segment the first battery cell image to be detected to obtain the first mask image, and the first battery cell image is detected according to the first mask image to obtain the glue detection result. In this way, on the one hand, the image segmentation model to be trained according to different film types can improve the efficiency of model training, and make the trained image segmentation model compatible with different types of film types; on the other hand, when the second mask image does not meet the accuracy condition, the image segmentation model is retrained by changing the film type, so that the image segmentation model obtained after training is more accurate, thereby improving the accuracy of glue detection.

[0074] In some embodiments, after acquiring the first battery cell image to be detected, the method may further include the following step S111:

[0075] Step S111: resetting the size of the first battery cell image to a first size, and filling the pixel values ​​of the first battery cell image after the size reset, so as to expand the size of the first battery cell image to a second size; wherein the second size is larger than the first size; and the second size is the size required for the trained image segmentation model to segment the first battery cell image.

[0076] In some implementations, the accuracy of the first battery cell image and the trained image segmentation model needs to be kept consistent, and therefore, the size of the first battery cell image needs to be reset.

[0077] In some embodiments, in order to maintain the ratio of the height and width of the first battery cell image, the size of the first battery cell image is first reset to a first size, where the first size is smaller than a second size required by the trained image segmentation model, and then the size of the first battery cell image is expanded from the first size to the second size by padding with zeros.

[0078] In some embodiments, the pixel values ​​of the first cell image after the size reset may be filled based on the neighboring pixel values. For example, the pixel values ​​to be filled may be determined based on the pixel values ​​of the surrounding pixels. Common neighboring pixel value selections include four neighborhoods and eight neighborhoods. The former only considers the pixel values ​​in four directions of up, down, left, and right, while the latter considers the pixel values ​​in eight directions of up, down, left, right, and diagonal.

[0079] The step S102 described above of using the trained image segmentation model to segment the glue-attached area in the first battery cell image to obtain a first segmentation result may include the following step S112:

[0080] Step S112: using the trained image segmentation model, segmenting the glue-attached area in the first battery cell image after the size expansion to obtain a first segmentation result.

[0081] In some embodiments, the input image of the trained image segmentation model is an image obtained by resizing the first battery cell image.

[0082] In the disclosed embodiment, the size of the acquired first battery cell image is reset to the first size, and the size of the first battery cell image is expanded to the second size required by the trained image segmentation model to segment the first battery cell image by pixel value filling, and the trained image segmentation model is used to segment the glue area in the first battery cell image after the size expansion. In this way, the size of the first battery cell image is reset to the first size, and then the first size is expanded to the second size by pixel value filling. This processing method can maintain the ratio of the height and width of the first battery cell image and prevent the loss of image accuracy.

[0083] The embodiment of the present disclosure provides a method for detecting adhesive sticking. Figure 2 A schematic diagram of the implementation process of a glue detection method provided in the embodiment of the present disclosure Figure 2 ,like Figure 2 As shown, the method includes the following steps S201 to S204:

[0084] Step S101: Acquire a first battery cell image of a battery cell to be detected;

[0085] Step S202: using the trained image segmentation model, performing target detection on the glue-attached area in the first battery cell image to obtain a target detection result, wherein the target detection result includes a target detection frame corresponding to the glue-attached area;

[0086] In some embodiments, the purpose of performing target detection on the glue-attached area in the first battery cell image is to determine the position of the glue-attached area.

[0087] In some implementations, the target detection frame is a detection frame selected from multiple candidate detection frames and having the highest probability of containing the glue area position.

[0088] Step S203: segmenting the glued area in the first battery cell image based on the target detection frame, and determining a first mask image corresponding to the glued area;

[0089] In some implementations, the first mask image is a binary image that can reflect the position information and shape information of the glue-applied area.

[0090] In some embodiments, the target position of the glue area in the first battery cell image can be determined according to the target detection frame; the pixel points at the target position and the pixel points outside the target position are set to different pixel values ​​to obtain a first mask image corresponding to the glue area.

[0091] Step S103: Based on the first mask image, perform glue detection on the first battery cell image to obtain a glue detection result.

[0092] In the disclosed embodiment, the trained image segmentation model is used to perform target detection on the glue area in the first battery cell image to obtain a target detection frame corresponding to the glue area, and based on the target detection frame, the glue area in the first battery cell image is segmented to obtain a first mask image corresponding to the glue area. In this way, the glue area is segmented according to the target detection frame obtained by target detection to obtain the first mask image; the target position of the glue area and the shape of the glue area can be obtained according to the first mask image.

[0093] In some embodiments, the step S202 described above of using the trained image segmentation model to perform target detection on the glued area in the first battery cell image to obtain the target detection result may include the following steps S211 and S212:

[0094] Step S211: using the trained image segmentation model, performing target detection on the glue-attached area in the first battery cell image to obtain at least one candidate detection frame corresponding to the glue-attached area;

[0095] In some embodiments, the output of target detection in the trained image segmentation model includes multiple candidate detection boxes and confidence levels corresponding to the multiple candidate detection boxes.

[0096] Step S212: Based on the at least one candidate detection frame, determine a target detection frame corresponding to the glue area.

[0097] In some implementations, a target detection box is selected from multiple candidate detection boxes.

[0098] In some implementations, a union of multiple candidate detection boxes is taken to obtain a target detection box.

[0099] It can be understood that, according to the confidences corresponding to multiple candidate detection frames, the candidate detection frames whose confidences are less than or equal to a preset threshold are removed from the multiple candidate detection frames, and non-maximum suppression is performed on the retained candidate detection frames to remove duplicate candidate detection frames. In this way, a more accurate target detection frame can be obtained.

[0100] In the disclosed embodiment, a target detection frame is determined from multiple candidate detection frames obtained by performing target detection on the glue-sticking area in the first battery cell image using a trained image segmentation model. In this way, a target detection frame is determined from multiple candidate detection frames obtained by target detection, and the target detection frame can accurately locate the target position of the glue-sticking area, making the detection result of the glue-sticking detection more accurate.

[0101] In some embodiments, the step S212 described above of determining the target detection frame corresponding to the glue area based on the at least one candidate detection frame may include the following steps S221 to S223:

[0102] Step S221: determining the confidence level corresponding to each of the candidate detection frames;

[0103] In some embodiments, the confidence level may represent the probability that the candidate detection box contains the glue area.

[0104] Step S222: selecting at least one first target detection frame from each of the candidate detection frames based on the confidence level corresponding to each of the candidate detection frames;

[0105] In some implementations, candidate detection frames corresponding to confidence levels greater than a preset threshold are retained, and candidate detection frames corresponding to confidence levels less than or equal to the preset threshold are removed, to obtain at least one first target detection frame.

[0106] Step S223: performing non-maximum suppression on each of the first target detection frames to determine the target detection frame.

[0107] In some implementations, non-maximum suppression is performed on multiple first target detection frames, and repeated first target detection frames are removed to obtain a target detection frame.

[0108] In the disclosed embodiment, first, the candidate detection frames corresponding to the confidences greater than the preset threshold are selected from each candidate detection frame, and then non-maximum suppression is performed on the selected candidate detection frames to obtain the target detection frame. In this way, the first target detection frame with a greater possibility is selected according to the confidences corresponding to each candidate detection frame; non-maximum suppression is performed on the first target detection frame to remove duplicate first target detection frames, and the final target detection frame is obtained.

[0109] In some embodiments, the candidate types include a first film type of the film attached to the front of the battery cell and a second film type of the film attached to the back of the battery cell. Determining a target film type different from the predicted film type from at least one candidate type in the above step S123 may include at least one of the following steps S151 and S152:

[0110] Step S151: when the predicted film type is the first film type, determining the target film type is the second film type;

[0111] Step S152: when the predicted film type is the second film type, determining the target film type to be the first film type.

[0112] In some embodiments, the film pasted on the front side of the battery cell is a single-color all-blue glue; and the film pasted on the back side of the battery cell is a combination of blue glue and transparent glue.

[0113] In some embodiments, the film attached to the front side of the battery cell and the film attached to the back side of the battery cell are all blue glue.

[0114] In some embodiments, the target film type is different from the predicted film type.

[0115] In the disclosed embodiment, when the predicted film type is the first film type, the target film type is determined to be the second film type; when the predicted film type is the second film type, the target film type is determined to be the first film type. In this way, by changing the film type and retraining the image segmentation model to be trained, a more accurate trained image segmentation model is obtained, thereby improving the accuracy of adhesive detection using the trained image segmentation model.

[0116] In some embodiments, the film affixed to the front side of the battery cell includes a film of a single color, and the film affixed to the back side of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color; or, the film affixed to the front side of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color, and the film affixed to the back side of the battery cell includes a film of a single color.

[0117] In some embodiments, the single color may be blue glue; the first film area of ​​the first color may be a film area of ​​blue glue, and the second film area of ​​the second color may be a film area of ​​transparent glue.

[0118] In some embodiments, the film attached to the front side of the battery cell and the film attached to the back side of the battery cell have different colors.

[0119] In the disclosed embodiment, if the film pasted on the front of the battery cell is a film of a single color, then the film pasted on the back includes a first film area of ​​a first color and a second film area of ​​a second color; or, if the film pasted on the front of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color, then the film pasted on the back of the battery cell includes a film of a single color. In this way, the film colors of the film pasted on the front of the battery cell and the film pasted on the back of the battery cell are different, so that the features of the film pasted on the front of the battery cell and the film pasted on the back of the battery cell are obviously different, so that the image segmentation model to be trained is trained using the second battery cell image and the target film type, which can improve the recognition ability of the trained image segmentation model for the film pasted on the front of the battery cell and the film pasted on the back of the battery cell.

[0120] In some embodiments, the second cell image has a glue area label. The above method may further include the following step S161:

[0121] Step S161: when the second mask image does not match the glue area label, determining that the second mask image does not meet the accuracy condition.

[0122] In some embodiments, the glue area label may indicate the target position of the glue area in the second battery cell image; and the second mask image may indicate the target position of the glue area after segmentation using the image segmentation model to be trained.

[0123] In some embodiments, by comparing the target position of the glued area in the second mask image with the target position of the glued area indicated by the glued area label, it can be determined whether the second mask image meets the accuracy condition.

[0124] In the embodiment of the present disclosure, when the second mask image does not match the label of the glued area, it is determined that the second mask image does not meet the accuracy condition. In this way, the label of the glued area is used as reference information, and the target position of the glued area in the second mask image is compared with the reference information to determine whether the second mask image meets the accuracy condition. This method can accurately determine whether the second mask image meets the accuracy condition.

[0125] In some embodiments, the above method may further include the following steps S171 and S172:

[0126] Step S171: obtaining a determination result in response to the accuracy determination operation on the second mask image;

[0127] In some implementations, it may be manually determined whether the second mask image meets the accuracy condition.

[0128] In some embodiments, the position and shape of the glue area of ​​the battery cell included in the second mask image obtained using the image segmentation model to be trained, and the position and shape of the glue area of ​​the battery cell included in the product information of the second battery cell image can be used to perform an accuracy determination operation on the position and shape of the glue area included in the second mask image based on the position and shape of the glue area of ​​the battery cell included in the product information of the second battery cell image, that is, determine the similarity between the position and shape of the glue area included in the second mask image and the position and shape of the glue area of ​​the battery cell included in the product information of the second battery cell image, and obtain a determination result.

[0129] In some implementations, the determination result may directly determine whether the second mask image satisfies the accuracy condition.

[0130] Step S172: Based on the determination result, determine whether the second mask image meets the accuracy condition.

[0131] In the embodiment of the present disclosure, the accuracy of the second mask image is determined to determine whether the second mask image meets the accuracy condition. In this way, by determining the accuracy of the second mask image, it is determined whether the second mask image meets the accuracy condition. This method can quickly determine whether the second mask image meets the accuracy condition.

[0132] In some embodiments, the above method may further include the following step S181:

[0133] Step S181: Acquire film color information of the first battery cell image; the film color information includes at least two colors or a single color.

[0134] The step S102 described above of using the trained image segmentation model to segment the glue-attached area in the first battery cell image to obtain a first segmentation result may include at least one of the following steps S182 and S183:

[0135] Step S182: when the film color information is at least two colors, using the trained image segmentation model, segmenting the glue-attached area in the first battery cell image to obtain a first segmentation result;

[0136] In some embodiments, the at least two colors may include blue glue and clear glue.

[0137] In some embodiments, if the film color information includes blue glue and transparent glue, the glue-applied area in the first battery cell image is segmented using a trained image segmentation model.

[0138] Step S183: when the film color information is a single color, segment the glued area in the first battery cell image using a color channel corresponding to the single color to obtain a first segmentation result.

[0139] In some embodiments, the single color is blue, that is, the adhesive area is entirely blue adhesive.

[0140] In some embodiments, if the film color information is full blue glue, the glue-applied area in the first battery cell image is segmented using the blue channel.

[0141] In the embodiment of the present disclosure, when the film color information is at least two colors, the trained image segmentation model is used to segment the glue area in the first battery cell image; when the film color information is a single color, the blue channel is used to segment the glue area in the first battery cell image. In this way, the blue channel is used to directly segment the glue area of ​​a single color, which is simpler and faster; the trained image segmentation model is used to segment the glue area of ​​at least two colors, which is compatible with glue areas of multiple colors, and the first segmentation result obtained is more accurate.

[0142] In some embodiments, the above step S103 may include the following steps S191 and S192:

[0143] Step S191: according to the first mask image, setting the glued area in the first battery cell image as a target pixel value to obtain a third battery cell image;

[0144] In some implementations, the target position of the adhesive area can be determined from the first mask image, and the target position is set to a target pixel value, generally set to blue, so that the transparent adhesive can be turned blue for easy visual detection.

[0145] Step S192: Based on the third battery cell image, perform a glue sticking test on the battery cell to be tested to obtain a glue sticking test result.

[0146] Through the glue area shown in the third battery cell image, it can be very clearly observed whether the glue area is accurate and complete.

[0147] In the disclosed embodiment, according to the first mask image, the glue area in the first battery cell image is set as the target pixel value to obtain the third battery cell image; according to the third battery cell image, the glue area of ​​the battery cell to be detected is detected to obtain the detection result of the glue area. In this way, according to the first mask image, the pixel value of the glue area is set as the target pixel value, which can better distinguish the glue area from other backgrounds, making the detection result of the glue area more accurate.

[0148] The following describes the application of the embodiments of the present disclosure in actual scenarios.

[0149] In the lithium battery production process, if the glue area of ​​the battery cell has abnormal coverage and does not cover the weld mark, it will cause the problem of leakage ear. Therefore, it is necessary to segment the glue area of ​​the battery cell to detect whether the glue area has abnormal coverage. In the related technology, the outline and coverage of the glue area of ​​the battery cell can be effectively extracted by adjusting the threshold of the blue channel or enhancing the contrast of the blue channel, and then it can be detected whether the glue pre-taken of the battery cell has abnormal coverage. However, due to changes in factors such as the lighting conditions and equipment status of the production line, it is often necessary to modify the threshold of the blue channel. To address this problem, Unet network segmentation is used to segment the glue area of ​​the battery cell, but this method often has insufficient segmentation accuracy and is prone to overkill.

[0150] On this basis, an embodiment of the present disclosure provides a method for dividing the adhesive laminating area of ​​a battery cell. Figure 3 A schematic diagram of the implementation flow of a method for segmenting the adhesive bonding area of ​​a battery cell provided in an embodiment of the present disclosure, such as Figure 3 As shown, the method may include the following steps S301 to S307:

[0151] Step S301: obtaining a glue-attached image of a battery cell to be inspected;

[0152] Use a camera or software to obtain the glue image of the battery cell to be tested.

[0153] Step S302: Determine whether to use the blue channel to directly detect the glue-attached area of ​​the battery cell; if so, proceed to step S303; if not, proceed to step S304;

[0154] If the adhesive image of the battery cell to be inspected has a single-color all-blue adhesive, and the adhesive has no wrinkles and the adhesive image is not overexposed, the blue channel is directly used to detect the adhesive of the battery cell; otherwise, the Yolov5-seg model is used to segment the adhesive of the battery cell.

[0155] Step S303: directly detecting the glue-attached area of ​​the battery cell using the blue channel;

[0156] By adjusting the threshold of the blue channel or enhancing the contrast of the blue channel, the outline and coverage of the adhesive of the battery cell can be effectively extracted, and then it can be detected whether the adhesive coverage of the battery cell is abnormal.

[0157] Step S304: segmenting the glue-attached area of ​​the battery cell using the Yolov5-seg model;

[0158] The Yolov5-seg model is used to segment the glue of the battery cell. Compared with the Unet network segmentation method, it can accurately segment the glue position of the battery cell. As the number of model training samples increases, it is compatible with the glue of various types of battery cells, reducing operation and maintenance costs. At the same time, the Yolov5-seg model is lighter and occupies less video memory.

[0159] For example, Figure 4 The glue image of the battery cell to be segmented using the Unet network segmentation method; Figure 5 The glue image of the battery cell to be segmented using the Yolov5-seg model; Figure 6 This is the segmentation result using the Unet network segmentation method; Figure 7 is the segmentation result using the Yolov5-seg model. Figure 4 The first adhesive area 401 and Figure 5 As shown in the second adhesive area 501, it can be seen that the second adhesive area 501 has wrinkles compared to the first adhesive area 401; Figure 6 and Figure 7 As shown, it can be seen Figure 7 The fourth glue area 701 of the split battery cell shown is more accurate and complete, and Figure 6 The third glue area 601 of the cell segmented as shown has a gap 602 in the middle, and the bottom of the third glue area 601 of the cell is incomplete, that is, the height of the glue area of ​​the segmented cell is not enough. Therefore, using the Yolov5-seg model to segment the glue area of ​​the cell can accurately segment the location of the glue area of ​​the cell.

[0160] The specific process of segmenting the adhesive of the battery cell using the Yolov5-seg model may include the following steps S311 to S316:

[0161] Step S311: obtaining a glue-attached image of the battery cell to be inspected, that is, an original image;

[0162] Step S312: Data set preparation;

[0163] Set the category of the data set, and set the labels to two categories: back side of glue (BM) and front side of glue (ZM). Front side of glue and back side of glue refer to the front and back sides of the battery cell. Figure 8 The schematic diagram of the film attached to the front of the battery cell is shown in FIG. Figure 8 As shown, the film pasted on the front is blue adhesive 801; Fig. 9 The schematic diagram of the film attached to the back of the battery cell is shown in FIG. Fig. 9 As shown, the films attached to the back include both blue adhesive 901 and transparent adhesive 902.

[0164] In the model training, methods such as scaling, rotating, shading, etc. are added to augment the images, so that good detection results can be obtained for transparent adhesives in different forms. Classification training is performed according to the set labels, and the augmentation processing is used to increase the samples of scenes such as overexposure, dirt, and spots.

[0165] Step S313: Yolov5-seg model preprocessing;

[0166] Initialize the Yolov5-seg model and set the detection accuracy of the Yolov5-seg model to 640*640.

[0167] Step S314: pre-processing the adhesive image of the battery cell to be inspected;

[0168] The size of the glue image of the battery cell to be tested is 3072*2048. In order to reset the size of the glue image of the battery cell to be tested to a size consistent with the detection accuracy of the Yolov5-seg model (i.e., 640*640), the size of the glue image of the battery cell to be tested is first reset to 640*512, and then the 640*512 size of the glue image of the battery cell to be tested is expanded to 640*640 by padding 0, and the expanded width and height values ​​are recorded.

[0169] Step S315: model reasoning;

[0170] The Yolov5-seg model contains the output of target detection (output0) and the output of segmentation (output1). Output0 usually contains the detection box, category confidence and category probability; output1 provides pixel-level classification information.

[0171] The preprocessed glue image of the battery cell to be tested is used as the input of the Yolov5-seg model. The Yolov5-seg model is parsed using optimization tools that can accelerate deep learning model reasoning (such as the tensorrt framework). All detection frames are obtained by obtaining the data output by output0. Some detection frames with low confidence are removed by the confidence corresponding to each detection frame, and some repeated detection frames are non-maximum suppressed. The target detection frame can be accurately screened to locate the coordinates of the glue area of ​​the battery cell. The segmented mask image can be obtained by the data output by output1. The segmented mask image is used to present the glue area of ​​the battery cell separately in a picture in a similar way to cutout.

[0172] Step S316: performing subsequent processing on the target detection results and segmentation results output by the Yolov5-seg model;

[0173] The target detection frame obtained by output0 in the Yolov5-seg model inference process is first used to complete the expanded width and height values ​​recorded in the preprocessing stage, and the target detection frame is precision-corrected. Then, the target detection frame is mapped back to the corresponding position in the original image 3072*2048 according to the scaling ratio to obtain the position of the glue area of ​​the battery cell in the original image; the mask image obtained by output1 is used to replace the part of the glue area of ​​the battery cell in the original image, and the segmented image is set to blue. In this way, the transparent glue part in the original image can be changed to the blue glue part, and then it can be used to locate whether the glue of the battery cell is pasted on the weld mark, and it can also be detected whether the glue of the battery cell is not pasted well and the tab is leaking.

[0174] When matching the adhesive of other types of batteries, even if the camera lens has a large area of ​​dirty carbon powder, the Yolov5-seg model can quickly give a high confidence score. For example, Fig.10 The adhesive image of the dirty battery cell to be tested is shown in FIG. Fig.10 As shown, there is a large area of ​​dirt 1001 in the adhesive image; Fig.11 The segmentation result of the adhesive image of the dirty battery cell to be inspected is shown in FIG. Fig.11 As shown, even if there is a large area of ​​dirt 1001 in the glue-sticking image of the battery cell to be inspected, the Yolov5-seg model can still accurately and completely segment the position of the glue-sticking area 1101 of the battery cell and the glue-sticking status.

[0175] Step S305: whether to segment the glue-attached area of ​​the battery cell; if so, proceed to step S306; if not, proceed to step S307;

[0176] Step S306: outputting a segmented image of the glue-attached area of ​​the battery cell;

[0177] Step S307: Output a segmentation failure flag.

[0178] The disclosed embodiment proposes a method for segmenting the glue area of ​​a battery cell based on the Yolov5-Seg model. The method combines deep learning technology to detect whether there is coverage anomaly in the glue area of ​​the battery cell, reducing the dependence on the configuration table parameters. In addition, there is no need to develop a set of pre-processing technologies separately for different imaging quality issues, so that the production line can be produced at high speed, which is convenient, fast and efficient. The method can accurately segment the position of the object, and as the model training samples increase, it is compatible with the detection of glue areas of various types of battery cells, reducing operation and maintenance costs. At the same time, the model used in this method is lighter and occupies less video memory.

[0179] Based on the foregoing embodiments, the embodiments of the present disclosure provide a glue sticking detection device, which includes the modules included, and the units included in the modules, etc., which can be implemented by a processor in a computer device; of course, it can also be implemented by a specific logic circuit; in the implementation process, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP) or a field programmable gate array (FPGA), etc.

[0180] Fig.12 A schematic diagram of the structure of a glue detection device provided in an embodiment of the present disclosure is shown in FIG. Fig.12 As shown, the adhesive detection device 1200 includes: a first acquisition module 1201, a first segmentation module 1202 and a detection module 1203, wherein:

[0181] A first acquisition module 1201 is used to acquire a first battery cell image of a battery cell to be detected;

[0182] The first segmentation module 1202 is used to segment the glue area in the first battery cell image using a trained image segmentation model to obtain a first segmentation result, wherein the first segmentation result includes a first mask image corresponding to the glue area; the trained image segmentation model is trained using the following model training method: obtaining a second battery cell image; using the image segmentation model to be trained, segmenting the glue area in the second battery cell image to obtain a second segmentation result, wherein the second segmentation result includes a second mask image corresponding to the glue area and a predicted film type; when the second mask image does not meet the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type; using the second battery cell image and the target film type, training the image segmentation model to be trained to obtain the trained image segmentation model;

[0183] The detection module 1203 is used to perform glue sticking detection on the first battery cell image based on the first mask image to obtain a glue sticking detection result.

[0184] In some embodiments, the glue detection device also includes: an expansion module, used to reset the size of the first battery cell image to a first size, and fill the pixel values ​​of the first battery cell image after the size reset, so as to expand the size of the first battery cell image to a second size; wherein the second size is larger than the first size; and the second size is the size required for the trained image segmentation model to segment the first battery cell image.

[0185] In some embodiments, the first segmentation module includes: a first segmentation unit, configured to segment the glue-attached area in the first battery cell image after size expansion using the trained image segmentation model to obtain a first segmentation result.

[0186] In some embodiments, the first segmentation module also includes: a target detection unit, used to use the trained image segmentation model to perform target detection on the glue area in the first battery cell image to obtain a target detection result, wherein the target detection result includes a target detection frame corresponding to the glue area; a first determination unit, used to segment the glue area in the first battery cell image based on the target detection frame, and determine a first mask image corresponding to the glue area.

[0187] In some embodiments, the target detection unit includes: a target detection subunit, used to use the trained image segmentation model to perform target detection on the glue area in the first battery cell image, and obtain at least one candidate detection frame corresponding to the glue area; a first determination subunit, used to determine the target detection frame corresponding to the glue area based on the at least one candidate detection frame.

[0188] In some embodiments, the first determination subunit is used to: determine the confidence corresponding to each of the candidate detection frames; based on the confidence corresponding to each of the candidate detection frames, filter out at least one first target detection frame from each of the candidate detection frames; perform non-maximum suppression on each of the first target detection frames to determine the target detection frame.

[0189] In some embodiments, the first segmentation module includes at least one of the following: a second determination unit for determining that the target film type is the second film type when the predicted film type is the first film type; and a third determination unit for determining that the target film type is the first film type when the predicted film type is the second film type.

[0190] In some embodiments, the film affixed to the front side of the battery cell includes a film of a single color, and the film affixed to the back side of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color; or, the film affixed to the front side of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color, and the film affixed to the back side of the battery cell includes a film of a single color.

[0191] In some embodiments, the glue detection device further includes: a first determination module, configured to determine that the second mask image does not meet an accuracy condition when the second mask image does not match the glue area label.

[0192] In some embodiments, the glue detection device further includes: a determination module for obtaining a determination result in response to an accuracy determination operation on the second mask image; and a second determination module for determining whether the second mask image meets an accuracy condition based on the determination result.

[0193] In some embodiments, the glue detection device also includes: a second acquisition module, used to acquire film color information of the first battery cell image; the film color information includes at least two colors or a single color; the first segmentation module includes at least one of the following: a second segmentation unit, used to segment the glue area in the first battery cell image using the trained image segmentation model when the film color information is at least two colors, and obtain a first segmentation result; a third segmentation unit, used to segment the glue area in the first battery cell image using a color channel corresponding to the single color when the film color information is a single color, and obtain a first segmentation result.

[0194] In some embodiments, the detection module includes: a setting submodule, used to set the glue area in the second battery cell image as the target pixel value according to the second mask image, so as to obtain a third battery cell image; and a detection submodule, used to perform glue detection on the battery cell to be detected based on the third battery cell image, so as to obtain a glue detection result.

[0195] The description of the above device embodiment is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. In some embodiments, the functions or modules included in the device provided in the embodiments of the present disclosure can be used to execute the method described in the above method embodiment. For technical details not disclosed in the device embodiment of the present disclosure, please refer to the description of the method embodiment of the present disclosure for understanding.

[0196] It should be noted that in the embodiments of the present disclosure, if the above method is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present disclosure is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a disk or an optical disk. In this way, the embodiments of the present disclosure are not limited to any specific hardware, software or firmware, or any combination of hardware, software, and firmware.

[0197] An embodiment of the present disclosure provides a computer device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the program, some or all of the steps in the above method are implemented.

[0198] The embodiment of the present disclosure provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, some or all of the steps in the above method are implemented. The computer-readable storage medium can be transient or non-transient.

[0199] An embodiment of the present disclosure provides a computer program, including a computer-readable code. When the computer-readable code is executed in a computer device, a processor in the computer device executes some or all of the steps for implementing the above method.

[0200] The embodiments of the present disclosure provide a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and when the computer program is read and executed by a computer, some or all of the steps in the above method are implemented. The computer program product can be implemented in hardware, software, or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium, and in other embodiments, the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc.

[0201] It should be noted here that the description of the various embodiments above tends to emphasize the differences between the various embodiments, and the same or similar aspects can be referenced to each other. The description of the above device, storage medium, computer program and computer program product embodiments is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment. For technical details not disclosed in the embodiments of the device, storage medium, computer program and computer program product disclosed in the present invention, please refer to the description of the method embodiment disclosed in the present invention for understanding.

[0202] It should be noted that Fig.13 is a schematic diagram of a hardware entity of a computer device in an embodiment of the present disclosure, such as Fig.13 As shown, the hardware entity of the computer device 1300 includes: a processor 1301, a communication interface 1302 and a memory 1303, wherein:

[0203] Processor 1301 generally controls the overall operation of computer device 1300 .

[0204] The communication interface 1302 enables the computer device to communicate with other terminals or servers through a network.

[0205] The memory 1303 is configured to store instructions and applications executable by the processor 1301, and can also cache data to be processed or processed by the processor 1301 and each module in the computer device 1300 (for example, image data, audio data, voice communication data, and video communication data), which can be implemented by flash memory (FLASH) or random access memory (Random Access Memory, RAM). Data can be transmitted between the processor 1301, the communication interface 1302, and the memory 1303 through the bus 1304.

[0206] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present disclosure. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in the various embodiments of the present disclosure, the size of the serial numbers of the above-mentioned steps / processes does not mean the order of execution. The execution order of each step / process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure. The serial numbers of the embodiments of the present disclosure are for description only and do not represent the advantages and disadvantages of the embodiments.

[0207] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0208] In the several embodiments provided in the present disclosure, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0209] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, the functional units in the various embodiments of the present disclosure may all be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0210] A person skilled in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, read-only memories (ROM), magnetic disks or optical disks.

[0211] Alternatively, if the above-mentioned integrated unit of the present disclosure is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure can essentially or in other words, the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0212] The above description is only an implementation mode of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed in the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A method for detecting adhesive bonding, characterized in that: The method comprises: Acquire a first battery cell image of a battery cell to be detected; Using the trained image segmentation model, the glue area in the first battery cell image is segmented to obtain a first segmentation result, and the first segmentation result includes a first mask image corresponding to the glue area; the trained image segmentation model is trained using the following model training method: obtaining a second battery cell image; using the image segmentation model to be trained, the glue area in the second battery cell image is segmented to obtain a second segmentation result, and the second segmentation result includes a second mask image corresponding to the glue area and a predicted film type; when the second mask image does not meet the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type; using the second battery cell image and the target film type, the image segmentation model to be trained is trained to obtain the trained image segmentation model; Based on the first mask image, performing glue detection on the first battery cell image to obtain a glue detection result; The at least one candidate type includes a first film type and a second film type; the determining of a target film type different from the predicted film type from the at least one candidate type includes at least one of the following: when the predicted film type is the first film type, determining the target film type to be the second film type; when the predicted film type is the second film type, determining the target film type to be the first film type; The using the second battery cell image and the target film type to train the image segmentation model to be trained includes: using the second battery cell image and the target film type as new training samples to train the image segmentation model to be trained; The second battery cell image has a glue area label; when the second mask image does not match the glue area label, determining that the second mask image does not meet the accuracy condition includes: when a first position corresponding to the second mask image is different from a second position corresponding to the glue area label, determining that the second mask image does not meet the accuracy condition.

2. The method for detecting adhesive bonding according to claim 1, characterized in that: After acquiring the first battery cell image of the battery cell to be detected, the method further includes: Resetting the size of the first battery cell image to a first size, and performing pixel value padding on the second battery cell image after the size reset, so as to expand the size of the first battery cell image to a second size; wherein the second size is larger than the first size; and the second size is the size required by the trained image segmentation model when segmenting the first battery cell image; The step of using the trained image segmentation model to segment the glue-attached area in the first battery cell image to obtain a first segmentation result includes: The trained image segmentation model is used to segment the glue-attached area in the first battery cell image after the size expansion to obtain a first segmentation result.

3. The method for detecting adhesive bonding according to claim 1, characterized in that: The step of using the trained image segmentation model to segment the glue-attached area in the first battery cell image to obtain a first segmentation result includes: Using the trained image segmentation model, perform target detection on the glue-attached area in the first battery cell image to obtain a target detection result, wherein the target detection result includes a target detection frame corresponding to the glue-attached area; Based on the target detection frame, the glued area in the first battery cell image is segmented to determine a first mask image corresponding to the glued area.

4. The method for detecting adhesive bonding according to claim 3, characterized in that: The using the trained image segmentation model to perform target detection on the glue-attached area in the first battery cell image to obtain a target detection result includes: Using the trained image segmentation model, perform target detection on the glue-attached area in the first battery cell image to obtain at least one candidate detection frame corresponding to the glue-attached area; Based on the at least one candidate detection frame, a target detection frame corresponding to the glue area is determined.

5. The method for detecting adhesive bonding according to claim 4, characterized in that: The determining, based on the at least one candidate detection frame, a target detection frame corresponding to the glue area includes: Determine the confidence level corresponding to each of the candidate detection frames; Based on the confidence scores corresponding to the candidate detection frames, at least one first target detection frame is selected from the candidate detection frames; Non-maximum suppression is performed on each of the first target detection frames to determine the target detection frame.

6. The method for detecting adhesive bonding according to any one of claims 1 to 5, characterized in that: The candidate types include a first film type of the film pasted on the front side of the battery cell and a second film type of the film pasted on the back side of the battery cell.

7. The method for detecting adhesive bonding according to claim 6, characterized in that: The film pasted on the front side of the battery cell includes a film of a single color, and the film pasted on the back side of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color; Alternatively, the film pasted on the front side of the battery cell includes a first film area of ​​a first color and a second film area of ​​a second color, and the film pasted on the back side of the battery cell includes a film of a single color.

8. The method for detecting adhesive bonding according to any one of claims 1 to 5, characterized in that: The method further comprises: In response to the accuracy determination operation on the second mask image, obtaining a determination result; Based on the determination result, it is determined whether the second mask image satisfies an accuracy condition.

9. The method for detecting adhesive bonding according to any one of claims 1 to 5, characterized in that: The method further comprises: Acquire film color information of the first battery cell image; the film color information includes at least two colors or a single color; The step of using the trained image segmentation model to segment the glue-attached area in the first battery cell image to obtain a first segmentation result includes at least one of the following: In a case where the film color information is at least two colors, using the trained image segmentation model, segmenting the glue-attached area in the first battery cell image to obtain a first segmentation result; In the case where the film color information is a single color, the glue-attached area in the first battery cell image is segmented using a color channel corresponding to the single color to obtain a first segmentation result.

10. The method for detecting adhesive bonding according to any one of claims 1 to 5, characterized in that: The step of performing glue detection on the first battery cell image based on the first mask image to obtain a glue detection result includes: According to the first mask image, setting the glued area in the first battery cell image as a target pixel value to obtain a third battery cell image; Based on the third battery cell image, a glue sticking detection is performed on the battery cell to be detected to obtain a glue sticking detection result.

11. A glue detection device, characterized in that: include: A first acquisition module, used to acquire a first battery cell image of a battery cell to be detected; A first segmentation module is used to segment the glue area in the first battery cell image using a trained image segmentation model to obtain a first segmentation result, wherein the first segmentation result includes a first mask image corresponding to the glue area; the trained image segmentation model is trained using the following model training method: obtaining a second battery cell image; using the image segmentation model to be trained, segmenting the glue area in the second battery cell image to obtain a second segmentation result, wherein the second segmentation result includes a second mask image corresponding to the glue area and a predicted film type; In the case where the second mask image does not satisfy the accuracy condition, determining a target film type different from the predicted film type from at least one candidate type; Using the second battery cell image and the target film type, training the image segmentation model to be trained to obtain the trained image segmentation model; a detection module, configured to perform a glue sticking detection on the first battery cell image based on the first mask image to obtain a glue sticking detection result; The at least one candidate type includes a first film type and a second film type; determining a target film type different from the predicted film type from the at least one candidate type includes: when the predicted film type is the first film type, determining the target film type to be the second film type; when the predicted film type is the second film type, determining the target film type to be the first film type; The using the second battery cell image and the target film type to train the image segmentation model to be trained includes: using the second battery cell image and the target film type as new training samples to train the image segmentation model to be trained; The second battery cell image has a glue area label; when the second mask image does not match the glue area label, determining that the second mask image does not meet the accuracy condition includes: when a first position corresponding to the second mask image is different from a second position corresponding to the glue area label, determining that the second mask image does not meet the accuracy condition.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, wherein: When the processor executes the program, the steps in the method according to any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 10 are implemented.

14. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps in the method according to any one of claims 1 to 10 are implemented.

Images