Battery insulation layer leakage detection method, device, equipment, medium and program product

Abstract

The embodiment of the present application provides a method, device, equipment, medium and program product for detecting the missing installation of the battery insulation layer. The battery insulation layer missing installation detection includes: obtaining a target image frame sequence, which is obtained by collecting images of the insulation layer installation station of the battery pack within a target time range; identifying the insulation layer installation status in each image frame in the target image frame sequence to obtain an installation status sequence; the installation status sequence includes the insulation layer installation status corresponding to multiple image frames; based on the installation status sequence, the insulation layer installation process within the target time range is detected to determine the detection result of the insulation layer missing installation event within the target time range. In this way, the detection rate of the insulation layer missing installation of the battery pack during the battery production process can be improved, the hardware computing cost of the detection can be reduced, and the workload of manual visual inspection can be reduced, thereby reducing labor costs.

Description

Technical Field

[0001] The present application relates to the field of battery production technology, and in particular to a method, device, equipment, medium and program product for detecting leakage of battery insulation layer. Background Art

[0002] This section is intended to provide a background or context for the embodiments of the present application. No description herein is admitted to be prior art by virtue of its inclusion in this section.

[0003] New energy batteries are being used more and more widely in life and industry. For example, new energy vehicles equipped with batteries have been widely used. In addition, batteries are also being increasingly used in areas such as energy storage.

[0004] In order to protect the battery safely, an insulating layer needs to be installed between the battery pack and the battery box cover during the battery production process. However, production operators may miss installing the insulating layer during the battery production process, resulting in an increased risk of fire in the battery product. In the related art, the insulating layer is checked to see if it is correctly installed by self-inspection by production staff and random inspection by patrol inspectors. However, this method is prone to missed inspections and has high labor costs. Summary of the invention

[0005] In view of this, the embodiments of the present application hope to provide a battery insulation layer missing detection method, device, equipment, medium and program product, which can improve the detection rate of battery pack insulation layer missing during battery production, reduce the hardware computing cost of detection, and reduce the workload of manual visual inspection, thereby reducing labor costs.

[0006] The technical solution of the embodiment of the present application is implemented as follows:

[0007] The present application embodiment provides a battery insulation layer missing detection method, comprising:

[0008] Acquire a target image frame sequence, where the target image frame sequence is obtained by collecting images of an insulation layer installation station of a battery pack within a target time range;

[0009] Identify the installation state of the insulation layer in each image frame in the target image frame sequence to obtain an installation state sequence; the installation state sequence includes the installation states of the insulation layer corresponding to the plurality of image frames;

[0010] Based on the installation state sequence, the insulation layer installation process within the target time range is detected to determine the detection result of the insulation layer missing event within the target time range.

[0011] In the battery insulation layer missing detection method of the embodiment of the present application, first, a target image frame sequence is obtained by performing image acquisition on the insulation layer installation station of the battery pack within the target time range; then, the insulation layer installation state in each image frame in the target image frame sequence is identified to obtain an installation state sequence; finally, based on the installation state sequence, the insulation layer installation process within the target time range is detected to determine the detection result of the insulation layer missing event within the target time range. In this way, since the insulation layer installation state in each image frame can more completely and directly reflect the insulation layer installation process, based on the installation state sequence, whether the insulation layer missing event occurs within the target time range can be more accurately detected, thereby improving the detection rate of the insulation layer missing of the battery pack during the battery production process; and, since the detection process does not require target tracking, the hardware computing cost of the detection is reduced, and the computing efficiency is improved. In addition, the battery insulation layer missing detection method of the embodiment of the present application can automatically detect whether the insulation layer missing event occurs within the target time range, thereby reducing the workload of manual visual inspection and reducing labor costs.

[0012] In some embodiments, the installation status of the insulation layer in each image frame in the target image frame sequence is identified to obtain an installation status sequence, including: for each image frame in the target image frame sequence, target detection is performed on the material located at the insulation layer installation station in the image frame to obtain the material detection result corresponding to the image frame, and based on the material detection result corresponding to the image frame, the installation status of the insulation layer corresponding to the image frame is determined; based on the insulation layer installation status corresponding to each image frame, the installation status sequence is determined.

[0013] In the above embodiment, by performing target detection on the material located at the insulating layer installation station in the image frame to determine the insulating layer installation status corresponding to the image frame, the insulating layer installation status recognition process can be simplified, so as to quickly and accurately obtain the installation status sequence, thereby improving the detection rate and detection efficiency of missing insulation layer of battery packs in the battery production process.

[0014] In some embodiments, the insulating layer installation station is used to install an insulating layer on a first material, and the insulating layer and the first material are used to be assembled with a second material to obtain a target battery product. The first material includes a battery pack or a battery box cover, and the second material includes one of the battery pack and the battery box cover that is different from the first material.

[0015] In the above embodiment, by installing the insulating layer on the battery pack or the battery box cover in the insulating layer installation station, it is convenient to assemble the insulating layer with the battery pack and the battery box cover to obtain the target battery product.

[0016] In some embodiments, based on the material detection result corresponding to the image frame, the installation status of the insulating layer corresponding to the image frame is determined, including: when the material detection result corresponding to the image frame includes a first detection result, determining that the installation status of the insulating layer corresponding to the image frame is that the first material has been put online; the first detection result indicates that the first material has been detected.

[0017] In the above embodiment, when the first material is detected, the installation state of the insulating layer corresponding to the image frame is determined to be that the first material has been put online. In this way, the installation state of the insulating layer corresponding to the image frame can be quickly and accurately determined according to the material detection result.

[0018] In some embodiments, based on the material detection result corresponding to the image frame, the installation status of the insulation layer corresponding to the image frame is determined, including: when the target detection result corresponding to the image frame includes a second detection result, determining that the installation status of the insulation layer corresponding to the image frame is the insulation layer being installed; the second detection result indicates that the insulation layer is detected to be in an installation state.

[0019] In the above embodiment, when an insulating layer in the installation state is detected, the insulating layer installation state corresponding to the image frame is determined to be the insulating layer installation state. In this way, the insulating layer installation state corresponding to the image frame can be quickly and accurately determined based on the material detection result.

[0020] In some embodiments, based on the material detection result corresponding to the image frame, the installation status of the insulation layer corresponding to the image frame is determined, including: when the target detection result corresponding to the image frame includes a third detection result, determining that the installation status of the insulation layer corresponding to the image frame is that the insulation layer has been installed; the third detection result indicates that the insulation layer is detected to be in a completed installation state.

[0021] In the above embodiment, when the insulating layer is detected to be in the installed state, the insulating layer installation state corresponding to the image frame is determined to be the insulating layer installed. In this way, the insulating layer installation state corresponding to the image frame can be quickly and accurately determined based on the material detection result.

[0022] In some embodiments, based on the material detection result corresponding to the image frame, the installation status of the insulating layer corresponding to the image frame is determined, including: when the target detection result corresponding to the image frame includes a fourth detection result, determining that the installation status of the insulating layer corresponding to the image frame is an idle state; the fourth detection result characterizes that the first material is not detected, the insulating layer is in an installation state, and the insulating layer is in an installation completed state.

[0023] In the above embodiment, when the first material, the insulating layer in the state of being installed, and the insulating layer in the state of being installed are not detected, the insulating layer installation state corresponding to the image frame is determined to be an idle state. In this way, the insulating layer installation state corresponding to the image frame can be quickly and accurately determined based on the material detection result.

[0024] In some embodiments, an installation state sequence is determined based on the insulation layer installation state corresponding to each image frame, including: determining a first candidate state sequence based on the insulation layer installation state corresponding to each image frame; performing noise reduction processing on the first candidate state sequence based on a sliding window of a target length to obtain a second candidate state sequence; and performing deduplication processing on adjacent insulation layer installation states in the second candidate state sequence to obtain an installation state sequence.

[0025] In the above embodiment, after determining the first candidate state sequence based on the insulation layer installation state corresponding to each image frame, a more accurate installation state sequence can be obtained through noise reduction and deduplication processing based on a sliding window, thereby reducing noise interference caused by inaccurate target detection, thereby further improving the detection rate of missing insulation layers in battery packs during battery production.

[0026] In some embodiments, the battery insulation layer missing detection method also includes: based on the product model of the target battery product, determining the operation time for a single installation of the insulation layer in the production process corresponding to the product model; based on the acquisition interval and operation time between adjacent image frames in the target image frame sequence, determining the target length.

[0027] In the above embodiment, on the one hand, based on the acquisition interval between adjacent image frames and the operation time of a single installation of the insulating layer, a more appropriate sliding window length can be determined, thereby improving the noise reduction effect on the first candidate state sequence; on the other hand, since the operation time of a single installation of the insulating layer is determined based on the product model of the target battery product, the determined sliding window length is more matched with the current production line, thereby further improving the noise reduction effect on the first candidate state sequence.

[0028] In some embodiments, based on the installation state sequence, the insulation layer installation process within the target time range is detected, including: detecting a candidate state subsequence from the installation state sequence, the candidate state subsequence uses an idle state as a starting state and an idle state as an ending state, and the starting state and the ending state include at least one of the following states: the first material has been put online, the insulation layer is being installed, and the insulation layer has been installed; when the candidate state subsequence is inconsistent with the target state sequence, it is determined that an insulation layer missing event occurs within the target time range.

[0029] In the above embodiment, using the idle state as the starting state and the ending state of the candidate state subsequence can more accurately locate the process of a single installation of the insulating layer, so that by comparing whether the candidate state subsequence is consistent with the target state sequence, it is possible to more accurately determine whether an insulating layer missing event occurs within the target time range, thereby improving the detection rate of missing insulation layer of the battery pack during the battery production process.

[0030] In some embodiments, target detection is performed on a material located at an insulation layer installation station in an image frame to obtain a material detection result corresponding to the image frame, including: utilizing a backbone network in a trained target detection model to extract features from the image frame to obtain multi-size first image features; utilizing a neck network in the target detection model to perform multi-scale feature fusion on the first image features to obtain second image features; utilizing a head network in the target detection model to predict a position and category of the material located at the insulation layer installation station in the image frame based on the second image features to obtain a material detection result corresponding to the image frame.

[0031] In the above embodiment, multi-scale first image features are extracted through the backbone network, and multi-scale feature fusion is performed on the first image features through the neck network. In this way, the fused second image features can contain more image frame features, so that the position and category of the material located at the insulation layer installation station in the image frame can be predicted based on the second image features, and more accurate material detection results can be obtained.

[0032] In some embodiments, the backbone network includes N layers of cascaded feature extraction modules, where N is a natural number greater than 2; the backbone network in the trained target detection model is used to extract features from image frames to obtain first image features of multiple sizes, including: performing convolution processing on the image frame to obtain the first input feature map; using the i-th layer of feature extraction module to perform the following operations: performing convolution processing on the i-th input feature map to obtain the i-th first intermediate feature map; performing convolution processing on the i-th first intermediate feature map after at least one residual connection processing to obtain the i-th second intermediate feature map; performing convolution processing on the i-th second intermediate feature map after at least one residual connection processing to obtain the i-th third intermediate feature map; performing convolution processing on the third intermediate feature map The method comprises the steps of: performing at least one residual connection operation on the i-th intermediate feature map to obtain the i-th fourth intermediate feature map; connecting the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map and the i-th fourth intermediate feature map by channel and then performing convolution processing to obtain the i-th output feature map; wherein, i is a natural number not exceeding N; performing pooling processing on the N-th output feature map to obtain the N-th output feature map after pooling; the first image feature includes the j-th output feature map and the N-th output feature map after pooling, and j is a natural number greater than 1 and less than N; the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map, the i-th fourth intermediate feature map and the i-th output feature map have the same size.

[0033] In the above embodiment, the feature extraction module in the backbone network enhances the information flow between multiple layers of convolution processing, making the gradient flow richer, and since the size of the output feature map in each layer of the feature extraction module is the same as the size of the first intermediate feature map, the second intermediate feature map, the third intermediate feature map, and the fourth intermediate feature map before the channel connection, that is, the channel scaling is increased by 4 times, more parameters are used for feature extraction overall, which can greatly improve the performance of the target detection model and improve the accuracy of target detection.

[0034] In some embodiments, the battery insulation layer missing detection method also includes: using a training image set to train a target detection model to be trained to obtain a trained target detection model; each training image in the training image set has a labeling box for marking the position of the detected target, and during the training process, positive and negative samples are allocated based on the intersection and union ratio between the predicted box and the labeled box of the detected target, and the distance from the center point of the labeled box to the predicted box, and the loss is calculated using the allocated positive and negative samples.

[0035] In the above embodiment, during the training of the target detection model, positive and negative samples are allocated based on the intersection-and-union ratio between the predicted box and the labeled box of the detected target, and the distance from the center point of the labeled box to the predicted box. This can increase the number of allocated positive samples and improve the detection accuracy of the trained target detection model.

[0036] In some embodiments, the insulating layer installation station includes an installation operation area corresponding to the product model of the battery product currently being produced; target detection is performed on the material located at the insulating layer installation station in the image frame to obtain the material detection result corresponding to the image frame, including: target detection is performed on the material located in the installation operation area in the image frame to obtain the material detection result corresponding to the image frame.

[0037] In the above embodiment, since the installation operation area corresponds to the product model of the battery product currently being produced, target detection of the material located in the installation operation area in the image frame can improve the accuracy of detection and greatly reduce the noise interference generated by target detection, thereby further improving the detection rate of missing insulation layers in battery packs during the battery production process.

[0038] The embodiment of the present application provides a battery insulation layer missing detection device, comprising:

[0039] A first acquisition module is used to acquire a target image frame sequence, where the target image frame sequence is obtained by acquiring images of an insulation layer installation station of a battery pack within a target time range;

[0040] An identification module is used to identify the installation state of the insulation layer in each image frame in the target image frame sequence to obtain an installation state sequence; the installation state sequence includes the installation states of the insulation layer corresponding to the multiple image frames respectively;

[0041] The detection module is used to detect the insulation layer installation process within the target time range based on the installation state sequence to determine the detection result of the insulation layer missing event within the target time range.

[0042] In the battery insulation layer missing detection device of the embodiment of the present application, since the insulation layer installation status in each image frame can more completely and directly reflect the insulation layer installation process, based on the installation status sequence, it is possible to more accurately detect whether the insulation layer missing event occurs within the target time range, thereby improving the detection rate of the battery pack insulation layer missing during the battery production process; and, since the detection process does not require target tracking, the hardware computing cost of the detection is reduced and the computing efficiency is improved. In addition, the battery insulation layer missing detection method of the embodiment of the present application can automatically detect whether the insulation layer missing event occurs within the target time range, thereby reducing the workload of manual visual inspection and reducing labor costs.

[0043] An embodiment of the present application 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, the steps in the above-mentioned battery insulation layer missing detection method are implemented.

[0044] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps in the above-mentioned battery insulation layer missing detection method are implemented.

[0045] An embodiment of the present application provides a computer program product, including a computer program or instructions. When the computer program or instructions are executed by a processor, the steps in the above-mentioned battery insulation layer missing detection method are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic diagram of the implementation process of a battery insulation layer missing detection method provided in an embodiment of the present application Figure 1 ;

[0047] Figure 2 A schematic diagram of a model structure of a target detection model provided in an embodiment of the present application;

[0048] Figure 3 A structural diagram of a feature extraction module provided in an embodiment of the present application;

[0049] Figure 4 A schematic diagram of the implementation process of a battery insulation layer missing detection method provided in an embodiment of the present application Figure 2 ;

[0050] Figure 5 A schematic diagram of an implementation architecture of a YOLO-C5 model training method provided in an embodiment of the present application;

[0051] Figure 6 A comparison of the effects of positive samples generated by a positive and negative sample allocation strategy provided in an embodiment of the present application and the positive and negative sample allocation strategy of YOLOv5 Figure 1 ;

[0052] Figure 7 A comparison of the effects of positive samples generated by a positive and negative sample allocation strategy provided in an embodiment of the present application and the positive and negative sample allocation strategy of YOLOv5 Figure 2 ;

[0053] Figure 8 A schematic diagram of performing sliding window denoising and adjacent state deduplication on a target sequence of video frames provided in an embodiment of the present application;

[0054] Fig. 9A schematic diagram of an implementation flow of retraining a target detection model provided in an embodiment of the present application;

[0055] Fig.10 A schematic diagram of the structure of a battery insulation layer missing detection device provided in an embodiment of the present application;

[0056] Fig.11 A schematic diagram of a hardware entity of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0057] It should be noted that, in the absence of conflict, the embodiments and technical features in the embodiments of the present application can be combined with each other, and the detailed description in the specific implementation method should be understood as an explanation of the purpose of the present application and should not be regarded as an improper limitation on the present application.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in this application are intended to cover non-exclusive inclusions.

[0059] In the description of the embodiments of the present application, the technical terms "first", "second", "third", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is more than two, unless otherwise clearly and specifically defined.

[0060] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0061] In the description of the embodiments of the present application, the term "and / or" is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0062] In the description of the embodiments of the present application, the orientation or position relationship indicated by technical terms such as "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "circumferential", "height direction", "first direction", and "second direction" are based on the orientation or position relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed, operated or used in a specific orientation. Therefore, they should not be understood as limitations on the embodiments of the present application.

[0063] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, technical terms such as "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.

[0064] In the description of the embodiments of the present application, unless otherwise clearly specified and limited, the technical term "contact" should be understood in a broad sense, and may be direct contact or contact through an intermediate medium layer. It may be contact with essentially no interaction force between the two contacting parties, or it may be contact with interaction force between the two contacting parties.

[0065] With the development of clean energy, more and more devices use electric energy as driving energy, and then power batteries that can store more electric energy and can be charged and discharged repeatedly are developing rapidly, such as lithium-ion batteries. Among them, power batteries are not only used in energy storage power supply systems such as hydropower, thermal power, wind power and solar power stations, but also widely used in electric vehicles such as electric bicycles, electric motorcycles, electric cars, as well as aerospace and other fields. With the continuous expansion of the application field of power batteries, the market demand is also constantly expanding.

[0066] In the embodiment of the present application, the battery pack can be made of battery cells and / or battery modules. A battery cell refers to a basic unit that can realize the mutual conversion of chemical energy and electrical energy, and can be used to make a battery module or a battery pack, so as to supply power to an electrical device. A battery cell can be a primary battery or a secondary battery. A secondary battery refers to a battery cell that can be activated by charging the active material after the battery cell is discharged and can continue to be used. The battery cell can be a lithium-ion battery, a sodium-ion battery, a sodium-lithium-ion battery, a lithium metal battery, a sodium metal battery, a lithium-sulfur battery, a magnesium-ion battery, a nickel-hydrogen battery, a nickel-cadmium battery, or a lead-acid battery, etc., and the embodiment of the present application is not limited to this. The battery cell can be cylindrical, rectangular, or in other shapes.

[0067] The battery cell includes an electrode assembly and an electrolyte. The electrode assembly is composed of a positive electrode sheet, a negative electrode sheet and a separator. The battery cell mainly relies on the movement of metal ions between the positive electrode sheet and the negative electrode sheet to work. The positive electrode sheet includes a positive electrode collector and a positive electrode active material layer. The positive electrode active material layer is coated on the surface of the positive electrode collector. The positive electrode collector not coated with the positive electrode active material layer protrudes from the positive electrode collector coated with the positive electrode active material layer. The positive electrode collector not coated with the positive electrode active material layer serves as the positive electrode tab. Taking lithium-ion batteries as an example, the material of the positive electrode collector can be aluminum, and the positive electrode active material can be lithium cobalt oxide, lithium iron phosphate, ternary lithium or lithium manganese oxide, etc. The negative electrode sheet includes a negative electrode collector and a negative electrode active material layer. The negative electrode active material layer is coated on the surface of the negative electrode collector. The negative electrode collector not coated with the negative electrode active material layer protrudes from the negative electrode collector coated with the negative electrode active material layer. The negative electrode collector not coated with the negative electrode active material layer serves as the negative electrode tab. The negative electrode current collector can be made of copper, and the negative electrode active material can be carbon or silicon. In order to ensure that a large current can pass without melting, the positive electrode tabs are multiple and stacked together, and the negative electrode tabs are multiple and stacked together. The material of the separator can be PP (Polypropylene) or PE (Polyethylene).

[0068] During the production process of batteries, in order to protect the batteries safely, it is necessary to install an insulating layer between the battery pack and the battery box cover. However, production operators may miss installing the insulating layer during the battery production process, resulting in an increased risk of fire in battery products. In the related art, the insulation layer is checked to see if it is installed correctly by means of self-inspection by production staff and random inspection by patrol inspectors. Self-inspection by production staff refers to the process in which production operators conduct self-inspection and self-verification of the battery packs they are responsible for during the production process to ensure that the insulation layer is correctly installed on the battery packs. Random inspection by patrol inspectors refers to the process in which special patrol inspectors randomly check whether the insulation layer is correctly installed on the battery packs on the production line during the production process. It can be seen that the methods of self-inspection by production staff and random inspection by patrol inspectors mainly rely on manual inspection, which is prone to missed inspections and has high labor costs.

[0069] The embodiment of the present application provides a method for detecting the missing of the battery insulation layer, which can be executed by a computer device. During implementation, the computer device may include a controller in a battery production system, such as an industrial computer, a programmable logic controller (PLC), a host computer, etc. The host computer may include but is not limited to at least one of a laptop computer, a tablet computer, a desktop computer, a large-screen device, a server, a mobile device, etc.

[0070] Figure 1 A schematic diagram of the implementation process of a battery insulation layer missing detection method provided in an embodiment of the present application Figure 1 ,like Figure 1 As shown, the battery insulation layer missing detection method includes the following steps S101 to S103:

[0071] Step S101, acquiring a target image frame sequence, where the target image frame sequence is obtained by capturing images of an insulation layer installation station of a battery pack within a target time range.

[0072] Here, the insulating layer installation station refers to a station in the battery production line for installing the insulating layer of the battery pack. During implementation, a suitable insulating layer installation station can be set according to the battery production process in the actual production line, and the embodiment of the present application is not limited to this.

[0073] An image acquisition device may be provided in the insulation layer installation station to acquire images of the insulation layer installation station of the battery pack. In some embodiments, the image acquisition device may sequentially acquire multiple image frames at a set acquisition interval.

[0074] In some implementations, the target time range refers to a time interval including the current time, and the length of the current time interval may be set according to actual conditions.

[0075] In some embodiments, the target time range may include multiple image acquisition time points forward and / or backward from the current moment, and the target image frame sequence may include image frames acquired by acquiring images of the insulation layer installation station of the battery pack at each image acquisition time point within the target time range. The number of image acquisition time points included in the target time range may be set according to actual conditions.

[0076] Step S102 , identifying the installation state of the insulation layer in each image frame in the target image frame sequence to obtain an installation state sequence; the installation state sequence includes the installation states of the insulation layer corresponding to the plurality of image frames.

[0077] Here, for the image frames in the target image frame sequence, the installation status of the insulating layer in the image frame can be identified to obtain the installation status of the insulating layer corresponding to the image frame. According to the installation status of the insulating layer corresponding to each image frame in the target image frame sequence, the installation status sequence corresponding to the target image frame sequence can be obtained.

[0078] In some embodiments, the material in the image frame can be detected, and the installation state of the insulating layer corresponding to the image frame can be determined based on whether the image frame contains the material and / or the type of material contained. For example, the correspondence between at least one material and / or material form and the installation state of the insulating layer can be preset, and the installation state of the insulating layer corresponding to the image frame can be determined based on the material and / or material form contained in the image frame and the correspondence. During implementation, the correspondence between each material and / or material form and the installation state of the insulating layer can be determined based on the production process of the battery product in the actual production scenario, and the embodiments of the present application are not limited to this.

[0079] In some embodiments, the action of the operator in the image frame can be identified to determine the action and / or state of the operator in the image frame, thereby determining the installation state of the insulating layer corresponding to the image frame according to the action and / or state of the operator. During implementation, those skilled in the art can use any appropriate action recognition model according to actual conditions to identify any action in the image frame, and the embodiments of the present application are not limited to this.

[0080] In some embodiments, the installation state sequence corresponding to the target image frame sequence may include the insulation layer installation state corresponding to each image frame in the target image frame sequence.

[0081] In some embodiments, the insulation layer installation status corresponding to each image frame in the target image frame sequence can be denoised and / or deduplicated to obtain an installation status sequence corresponding to the target image frame sequence, that is, the installation status sequence corresponding to the target image frame sequence can include the insulation layer installation status corresponding to some image frames in the target image frame sequence.

[0082] Step S103: Based on the installation state sequence, the insulation layer installation process within the target time range is detected to determine the detection result of the insulation layer missing event within the target time range.

[0083] Here, it can be understood that the installation state sequence corresponding to the target image frame sequence can reflect the insulation layer installation process within the target time range, so by analyzing and detecting the installation state sequence, it can be determined whether an insulation layer missing event occurs within the target time range.

[0084] In some embodiments, a target state sequence characterizing that the insulation layer has been correctly installed can be determined based on the installation operations involved in a standard insulation layer installation process. By comparing whether the installation state sequence corresponding to the target image frame sequence is consistent with the target state sequence, it can be determined whether an insulation layer missing event occurs within the target time range, that is, a detection result of an insulation layer missing event within the target time range is obtained.

[0085] In some embodiments, a target sequence of at least two insulation layer installation states that needs to be met during the insulation layer installation process can be determined based on the sequence of at least two installation operations in a standard insulation layer installation process. By comparing whether the sequence of at least two insulation layer installation states in an installation state sequence corresponding to a target image frame sequence meets the target sequence, it can be determined whether an insulation layer missing event occurs within a target time range.

[0086] In the battery insulation layer missing detection method of the embodiment of the present application, first, a target image frame sequence is obtained by performing image acquisition on the insulation layer installation station of the battery pack within the target time range; then, the insulation layer installation state in each image frame in the target image frame sequence is identified to obtain an installation state sequence; finally, based on the installation state sequence, the insulation layer installation process within the target time range is detected to determine the detection result of the insulation layer missing event within the target time range. In this way, since the insulation layer installation state in each image frame can more completely and directly reflect the insulation layer installation process, based on the installation state sequence, whether the insulation layer missing event occurs within the target time range can be more accurately detected, thereby improving the detection rate of the insulation layer missing of the battery pack during the battery production process; and, since the detection process does not require target tracking, the hardware computing cost of the detection is reduced, and the computing efficiency is improved. In addition, the battery insulation layer missing detection method of the embodiment of the present application can automatically detect whether the insulation layer missing event occurs within the target time range, thereby reducing the workload of manual visual inspection and reducing labor costs.

[0087] In some embodiments, the above step S102 may include the following steps S111 to S112:

[0088] Step S111, for each image frame in the target image frame sequence, target detection is performed on the material located at the insulation layer installation station in the image frame to obtain the material detection result corresponding to the image frame, and based on the material detection result corresponding to the image frame, the insulation layer installation status corresponding to the image frame is determined.

[0089] Here, any suitable target detection model may be used to detect the object in the image frame, and the present application embodiment is not limited to this. For example, the target detection model may include but is not limited to at least one of a Regions with CNN features (R-CNN) model, a YOLO (You Only Look Once) model, and the like.

[0090] In some embodiments, the material detection result may include whether the target material is detected and / or the type of the detected material. The target material may include but is not limited to at least one of the following types of materials: a battery pack, an insulating layer, and a battery box cover.

[0091] Step S112, determining an installation state sequence based on the installation state of the insulation layer corresponding to each image frame.

[0092] In some implementations, the installation state sequence may be obtained by arranging the installation states of the insulation layers corresponding to each image frame in sequence from front to back according to the acquisition time of each image frame.

[0093] In some embodiments, the installation state of the insulation layer corresponding to each image frame can be arranged in sequence from front to back according to the acquisition time of each image frame, and then noise reduction processing and / or deduplication processing can be performed to obtain an installation state sequence. For example, the noise reduction process may include a smooth noise reduction process based on a sliding window. For another example, the noise reduction process may include: when the confidence level of the material detection result corresponding to the image frame is less than a confidence threshold, the installation state of the insulation layer corresponding to the image frame is removed from the installation state sequence. During implementation, those skilled in the art may adopt any suitable noise reduction processing method according to actual conditions, and the embodiments of the present application are not limited to this.

[0094] In the above embodiment, by performing target detection on the material located at the insulating layer installation station in the image frame to determine the insulating layer installation status corresponding to the image frame, the insulating layer installation status recognition process can be simplified, so as to quickly and accurately obtain the installation status sequence, thereby improving the detection rate and detection efficiency of missing insulation layer of battery packs in the battery production process.

[0095] In some embodiments, the insulating layer installation station is used to install an insulating layer on a first material, and the insulating layer and the first material are used to be assembled with a second material to obtain a target battery product. The first material includes a battery pack or a battery box cover, and the second material includes one of the battery pack and the battery box cover that is different from the first material.

[0096] For example, the first material includes a battery pack, the second material includes a battery box cover, the insulation layer installation station is used to install an insulation layer on the battery pack, and the battery pack with the insulation layer installed is used to be assembled with the battery box cover at a downstream station to obtain a target battery product.

[0097] For another example, the first material includes a battery box cover, the second material includes a battery pack, the insulation layer installation station is used to install an insulation layer on the battery box cover, and the battery box cover with the insulation layer installed is used to be assembled with the battery pack at a downstream station to obtain a target battery product.

[0098] In the above embodiment, by installing the insulating layer on the battery pack or the battery box cover in the insulating layer installation station, it is convenient to assemble the insulating layer with the battery pack and the battery box cover to obtain the target battery product.

[0099] In some embodiments, determining the installation state of the insulating layer corresponding to the image frame based on the material detection result corresponding to the image frame in the above step S111 may include the following steps S121:

[0100] Step S121, when the material detection result corresponding to the image frame includes the first detection result, determining that the installation state of the insulation layer corresponding to the image frame is that the first material has been put online; the first detection result indicates that the first material has been detected.

[0101] In the above embodiment, when the first material is detected, the installation state of the insulating layer corresponding to the image frame is determined to be that the first material has been put online. In this way, the installation state of the insulating layer corresponding to the image frame can be quickly and accurately determined according to the material detection result.

[0102] In some embodiments, determining the installation state of the insulating layer corresponding to the image frame based on the material detection result corresponding to the image frame in the above step S111 may include the following step S122:

[0103] Step S122, when the target detection result corresponding to the image frame includes a second detection result, determining that the installation state of the insulation layer corresponding to the image frame is the insulation layer being installed; the second detection result indicates that the insulation layer is detected to be in the installation state.

[0104] In the above embodiment, when an insulating layer in the installation state is detected, the insulating layer installation state corresponding to the image frame is determined to be the insulating layer installation state. In this way, the insulating layer installation state corresponding to the image frame can be quickly and accurately determined based on the material detection result.

[0105] In some embodiments, determining the installation state of the insulating layer corresponding to the image frame based on the material detection result corresponding to the image frame in the above step S111 may include the following step S123:

[0106] Step S123, when the target detection result corresponding to the image frame includes the third detection result, determining that the installation state of the insulation layer corresponding to the image frame is that the insulation layer has been installed; the third detection result indicates that the insulation layer is detected to be in an installed state.

[0107] In the above embodiment, when the insulating layer is detected to be in the installed state, the insulating layer installation state corresponding to the image frame is determined to be the insulating layer installed. In this way, the insulating layer installation state corresponding to the image frame can be quickly and accurately determined based on the material detection result.

[0108] In some embodiments, determining the installation state of the insulating layer corresponding to the image frame based on the material detection result corresponding to the image frame in the above step S111 may include the following step S124:

[0109] Step S124, when the target detection result corresponding to the image frame includes the fourth detection result, determine that the installation state of the insulating layer corresponding to the image frame is an idle state; the fourth detection result indicates that the first material is not detected, the insulating layer is in the installation state, and the insulating layer is in the installation completed state.

[0110] In the above embodiment, when the first material, the insulating layer in the state of being installed, and the insulating layer in the state of being installed are not detected, the insulating layer installation state corresponding to the image frame is determined to be an idle state. In this way, the insulating layer installation state corresponding to the image frame can be quickly and accurately determined based on the material detection result.

[0111] In some embodiments, determining the installation status of the insulating layer corresponding to the image frame based on the material detection result corresponding to the image frame in the above step S111 may include at least one of the above steps S121 to S124.

[0112] In some embodiments, the above step S112 may include the following steps S131 to S133:

[0113] Step S131, determining a first candidate state sequence based on the insulating layer installation state corresponding to each image frame;

[0114] Step S132, based on the sliding window of the target length, performing noise reduction processing on the first candidate state sequence to obtain a second candidate state sequence;

[0115] Step S133, performing deduplication processing on adjacent insulating layer installation states in the second candidate state sequence to obtain an installation state sequence.

[0116] Here, the insulating layer installation state corresponding to each image frame may be arranged in sequence according to the order of the acquisition time of each image frame in the target image frame sequence from front to back, so as to obtain the first candidate state sequence.

[0117] The target length can be pre-set by a technician in this field according to actual conditions, or it can be determined based on the acquisition interval between adjacent image frames in the target image frame sequence, the operation time for a single installation of the insulating layer, and / or the speed at which the operator installs the insulating layer, etc. The embodiments of the present application are not limited to this.

[0118] During implementation, those skilled in the art may perform noise reduction processing on the first candidate state sequence in any suitable manner based on a sliding window of a target length according to actual conditions, and the embodiments of the present application are not limited thereto.

[0119] In some embodiments, based on a sliding window of a target length, a first candidate state sequence is subjected to denoising processing to obtain a second candidate state sequence, which may include: for the kth insulation layer installation state in the first candidate state sequence, taking the kth insulation layer installation state as the tail of the current sliding window, obtaining the insulation layer installation state that appears most times in the current sliding window, and replacing the kth insulation layer installation state with the insulation layer installation state that appears most times, wherein k is a natural number, k is greater than 0 and less than the length of the first candidate state sequence.

[0120] In some embodiments, based on a sliding window of a target length, a first candidate state sequence is subjected to denoising processing to obtain a second candidate state sequence, which may include: for the kth insulation layer installation state in the first candidate state sequence, taking the kth insulation layer installation state as the midpoint position of the current sliding window, if the first m1 insulation layer installation states of the kth insulation layer installation state in the current sliding window are the same, and the first m1 insulation layer installation states are different from the kth insulation layer installation state and are the same as the last m2 insulation layer installation states of the kth insulation layer installation state, then the kth insulation layer installation state is updated to be the same as the first m1 insulation layer installation states, wherein k is a natural number, k is greater than 0 and less than the length of the first candidate state sequence, and m1 and m2 are both natural numbers greater than 1.

[0121] In some embodiments, the process of deduplicating adjacent insulation layer installation states in the second candidate state sequence may include: when there are at least two adjacent and repeated insulation layer installation states in the second candidate state sequence, removing the second and subsequent insulation layer installation states of the at least two adjacent and repeated insulation layer installation states.

[0122] For example, the first candidate state sequence is: [idle state, idle state, idle state, first material online, first material online, first material online, insulation layer installing, insulation layer installing, insulation layer installing, insulation layer installed, insulation layer installed, insulation layer installed, insulation layer installed, insulation layer installing, insulation layer installed, insulation layer installed, insulation layer installed, idle state, idle state, idle state, idle state], where the insulation layer installing between insulation layer installed and insulation layer installed is noise (i.e., false detection result); using a sliding window of length 6, taking the insulation layer installing state with the most states in the window to smooth the first candidate state sequence can eliminate the noise , the second candidate state sequence is as follows: [idle state, idle state, idle state, idle state, idle state, idle state, idle state, the first material is online, the first material is online, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, the idle state]; then the adjacent insulation layer installation states in the second candidate state sequence are deduplicated, and the installation state sequence can be obtained as follows: [idle state, the first material is online, the insulation layer is installed, the insulation layer is installed, the insulation layer is installed, idle state].

[0123] In the above embodiment, after determining the first candidate state sequence based on the insulation layer installation state corresponding to each image frame, a more accurate installation state sequence can be obtained through noise reduction and deduplication processing based on a sliding window, thereby reducing noise interference caused by inaccurate target detection, thereby further improving the detection rate of missing insulation layers in battery packs during battery production.

[0124] In some embodiments, the above-mentioned battery insulation layer missing detection method may further include the following steps S141 to S142:

[0125] Step S141, based on the product model of the target battery product, determining the operation time of installing the insulation layer once in the production process corresponding to the product model;

[0126] Step S142, determining the target length based on the acquisition interval and operation time between adjacent image frames in the target image frame sequence.

[0127] Here, considering that the installation method, connection structure, etc. between the insulating layer and the battery cell and / or the battery box cover may be different in battery products of different product models, the production processes used may be different, and the process of installing the insulating layer may also be different. Therefore, the operation time for a single installation of the insulating layer may be different in the production process corresponding to different product models.

[0128] In some embodiments, the operation time of installing the insulating layer once for each of a plurality of product models may be preconfigured, and the corresponding operation time of installing the insulating layer once may be determined based on the product model of the target battery product.

[0129] In some embodiments, the production process corresponding to the product model of the target battery product may be determined, and the operation time for installing the insulating layer once may be determined based on the production process.

[0130] In some embodiments, the number of image frames collected during a single installation of the insulating layer can be determined based on the collection interval between adjacent image frames in the target image frame sequence and the operation time of a single installation of the insulating layer, and the target length of the sliding window can be determined based on the number. For example, the number of image frames collected during the single installation of the insulating layer can be directly determined as the target length of the sliding window. For another example, the number of image frames collected during the single installation of the insulating layer can be multiplied by a preset multiple and then rounded to obtain the target length of the sliding window, and the preset multiple can be greater than 0 and less than 1.

[0131] In the above embodiment, on the one hand, based on the acquisition interval between adjacent image frames and the operation time of a single installation of the insulating layer, a more appropriate sliding window length can be determined, thereby improving the noise reduction effect on the first candidate state sequence; on the other hand, since the operation time of a single installation of the insulating layer is determined based on the product model of the target battery product, the determined sliding window length is more matched with the current production line, thereby further improving the noise reduction effect on the first candidate state sequence.

[0132] In some embodiments, the above step S103 may include the following steps S151 to S152:

[0133] Step S151, detecting a candidate state subsequence from the installation state sequence, wherein the candidate state subsequence takes the idle state as the starting state and the idle state as the ending state, and the starting state and the ending state include at least one of the following states: the first material has been put online, the insulation layer is being installed, and the insulation layer has been installed;

[0134] Step S152: when the candidate state subsequence is inconsistent with the target state sequence, determining that an insulation layer missing event occurs within the target time range.

[0135] Here, the target state sequence may be an insulating layer installation state sequence that indicates that the insulating layer has been correctly installed. During implementation, those skilled in the art may determine a suitable target state sequence according to the actual production process, and the embodiments of the present application are not limited thereto.

[0136] In some embodiments, the target state sequence includes idle state, first material online, insulation layer being installed, insulation layer installed, idle state arranged in sequence, that is, [idle state, first material online, insulation layer being installed, insulation layer installed, idle state].

[0137] In the above embodiment, using the idle state as the starting state and the ending state of the candidate state subsequence can more accurately locate the process of a single installation of the insulating layer, so that by comparing whether the candidate state subsequence is consistent with the target state sequence, it is possible to more accurately determine whether an insulating layer missing event occurs within the target time range, thereby improving the detection rate of missing insulation layer of the battery pack during the battery production process.

[0138] In some embodiments, the above step S111 performs target detection on the material located at the insulating layer installation station in the image frame to obtain the material detection result corresponding to the image frame, which may include the following steps S161 to S163:

[0139] Step S161, using the backbone network in the trained object detection model to extract features from the image frame to obtain first image features of multiple sizes;

[0140] Step S162, using the neck network in the target detection model, performing multi-scale feature fusion on the first image feature to obtain a second image feature;

[0141] Step S163, using the head network in the target detection model, based on the second image feature, predict the position and category of the material located at the insulation layer installation station in the image frame to obtain the material detection result corresponding to the image frame.

[0142] During implementation, the backbone network, neck network and / or head network in the target detection model can be selected according to actual conditions, and the embodiments of the present application are not limited to this.

[0143] In some embodiments, the first image feature may include multiple image feature maps of different sizes; the multiple image feature maps of different sizes may be feature fused using the neck network to obtain the second image feature.

[0144] In some embodiments, the target detection model may include a YOLO model, such as a YOLOv4 model, a YOLOv5 model, and / or a YOLOv7 model.

[0145] In some implementations, the target detection model may be obtained by improving the feature extraction network in the YOLOv5 model. For example, the feature extraction module (such as the C3 module) in the YOLOv5 model may be improved to obtain the target detection model.

[0146] In some embodiments, the model structure of the target detection model can be as follows: Figure 2 As shown, the target detection model may include a backbone network 110, a neck network 120 and a head network 130, wherein Conv represents a convolutional layer, C5 represents a feature extraction module, PL represents a pooling layer, upsample represents an upsampling layer, Concat represents a channel connection layer, Conv2d represents a two-dimensional convolution, dim-trans represents a dimensionality transformation layer, and BBbox-Dec represents a decoding module.

[0147] In some implementations, a category score threshold filtering module, a confidence and category score product threshold filtering module, and / or a non-maximum suppression module may be included after the decoding module.

[0148] It should be noted that the target detection model is not limited to Figure 2 When implementing the model structure shown, those skilled in the art can determine any suitable model structure for the target detection model according to actual conditions, and the embodiments of the present application are not limited to this.

[0149] In the above embodiment, multi-scale first image features are extracted through the backbone network, and multi-scale feature fusion is performed on the first image features through the neck network. In this way, the fused second image features can contain more image frame features, so that the position and category of the material located at the insulation layer installation station in the image frame can be predicted based on the second image features, and more accurate material detection results can be obtained.

[0150] In some embodiments, the backbone network includes N layers of cascaded feature extraction modules, where N is a natural number greater than 2.

[0151] The above step S161 may include the following steps S171 to S173:

[0152] Step S171, performing convolution processing on the image frame to obtain the first input feature map;

[0153] Step S172, using the i-th layer feature extraction module to perform the following operations: performing convolution processing on the i-th input feature map to obtain the i-th first intermediate feature map; performing at least one residual connection processing on the i-th first intermediate feature map and then performing convolution processing to obtain the i-th second intermediate feature map; performing at least one residual connection processing on the i-th second intermediate feature map and then performing convolution processing to obtain the i-th third intermediate feature map; performing at least one residual connection operation on the third intermediate feature map to obtain the i-th fourth intermediate feature map; connecting the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map and the i-th fourth intermediate feature map according to the channel and then performing convolution processing to obtain the i-th output feature map; wherein i is a natural number not exceeding N;

[0154] Step S173, performing pooling processing on the Nth output feature map to obtain the Nth output feature map after pooling; the first image feature includes the jth output feature map and the Nth output feature map after pooling, where j is a natural number greater than 1 and less than N; the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map, the i-th fourth intermediate feature map and the i-th output feature map have the same size.

[0155] Here, the backbone network may include more than three layers of cascaded feature extraction modules, and each layer of feature extraction modules is used to extract features from the input feature map of the current layer, obtain the output feature map of the current layer, and output the output feature map of the current layer to the feature extraction module of the next layer.

[0156] For the last layer feature extraction module (i.e., the Nth layer feature extraction module), after obtaining the Nth output feature map, the Nth output feature map can be input into the pooling layer for pooling processing to obtain and output the pooled Nth output feature map.

[0157] During implementation, any suitable pooling method can be used to perform pooling processing on the Nth output feature map, and the embodiments of the present application are not limited to this.

[0158] In some implementations, the Nth output feature map may be subjected to a spatial pyramid pooling process to obtain a pooled Nth output feature map. For example, the Nth output feature map may be input into a spatial pyramid pooling layer to perform a spatial pyramid pooling process to obtain and output a pooled Nth output feature map.

[0159] It can be understood that, for the feature extraction module of the same layer, the first intermediate feature map, the second intermediate feature map, the third intermediate feature map, the fourth intermediate feature map and the output feature map calculated in the layer have the same size; the sizes of the first intermediate feature map / second intermediate feature map / third intermediate feature map / fourth intermediate feature map / output feature map of feature extraction modules of different layers may be the same or different.

[0160] In some embodiments, see Figure 3, the feature extraction module (hereinafter referred to as C5 module) includes the first convolution layer Conv1, the second convolution layer Conv2, the first bottleneck layer Bottleneck1, the third convolution layer Conv3, the second bottleneck layer Bottleneck2, the fourth convolution layer Conv4, the third bottleneck layer Bottleneck3, the connection layer Concat and the fifth convolution layer Conv5; the first convolution layer Conv1 and the second convolution layer Conv2 are both used to perform convolution processing on the input feature map and output the first intermediate feature map; the first bottleneck layer Bottleneck1 is used to perform at least one residual connection processing on the first intermediate feature map output by the second convolution layer Conv2 to obtain the first intermediate feature map after the residual connection processing, and the third convolution layer Conv3 is used to perform convolution processing on the first intermediate feature map after the residual connection processing to obtain the second intermediate feature map; the second bottleneck layer Bottleneck2 is used to perform at least one residual connection process on the second intermediate feature map to obtain the second intermediate feature map after the residual connection process. The fourth convolution layer Conv4 is used to perform convolution processing on the second intermediate feature map after the residual connection process to obtain the third intermediate feature map; the third bottleneck layer Bottleneck3 is used to perform at least one residual connection operation on the third intermediate feature map to obtain the fourth intermediate feature map; the connection layer Concat is used to connect the first intermediate feature map output by the first convolution layer Conv1, the second intermediate feature map output by the third convolution layer Conv3, the third intermediate feature map output by the fourth convolution layer Conv4, and the fourth intermediate feature map output by the third bottleneck layer Bottleneck3 by channel to obtain the connected intermediate feature map; the fifth convolution layer Conv5 is used to perform convolution processing on the connected intermediate feature map to obtain the output feature map. Among them, the implementation method of each convolution layer and bottleneck layer can be determined by those skilled in the art according to actual conditions.

[0161] In some embodiments, the bottleneck layer in the feature extraction module can implement a DarknetBottleneck function, which is a function with a residual connection.

[0162] For example, the input feature map of the feature extraction module can be denoted as X, the batch size during training is B, the dimension of the height of the input feature map is h, the dimension of the width of the input feature map is w, and the channel size of the input feature map is c_in. The convolution function used in each convolution layer is denoted as Cov, the size of the convolution kernel W is 1*1, the convolution kernel step is 1, the expansion edge is 0, and the output channel is c_out. The DarknetBottleneck function used in the bottleneck layer is denoted as f. The merge operation is denoted as Con (i.e., channel connection operation). The process of feature extraction using the feature extraction module is as follows:

[0163] 1) Input features Figure X Perform convolution processing to obtain the first intermediate feature map , The size is B*h*w*c_out, and the number of output channels is c_out;

[0164] 2) Input features Figure X After the convolution process, the residual connection process is performed n times, and then the convolution process is performed again to obtain the second intermediate feature map. , The size of is B*h*w*c_out, n is a positive integer;

[0165] 3) The second intermediate feature map feature map After performing n residual connection processing, convolution processing is performed to obtain the third intermediate feature map , The size is B*h*w*c_out;

[0166] 4) The third intermediate feature map Perform n residual connection processing to obtain the fourth intermediate feature map , The size is B*h*w*c_out;

[0167] 5) The feature map Perform a merge operation and then perform convolution to obtain the output feature map , the size of Y is B*h*w*c_out.

[0168] It is understandable that based on the feature map Get the output feature map In the process, a 4-fold channel scaling is performed to output the feature map Size and features of The same size.

[0169] It should be noted that in the target detection model provided in the embodiment of the present application, the feature extraction module of the backbone network is not limited to Figure 3 In the structure shown, when implementing, those skilled in the art can select any suitable feature extraction module for the backbone network according to the actual situation, and the embodiments of the present application are not limited to this. For example, the feature extraction module of the backbone network can also include at least one level of convolutional layer and bottleneck layer in series between the third bottleneck layer Bottleneck3 and the connection layer Concat (such as the sixth convolutional layer, the fourth bottleneck layer, the seventh convolutional layer, the fifth bottleneck layer... cascaded in sequence), and the feature map output by at least one convolutional layer is also input to the connection layer Concat.

[0170] In the above embodiment, the feature extraction module in the backbone network enhances the information flow between multiple layers of convolution processing, making the gradient flow richer, and since the size of the output feature map in each layer of the feature extraction module is the same as the size of the first intermediate feature map, the second intermediate feature map, the third intermediate feature map, and the fourth intermediate feature map before the channel connection, that is, the channel scaling is increased by 4 times, more parameters are used for feature extraction overall, which can greatly improve the performance of the target detection model and improve the accuracy of target detection.

[0171] In some embodiments, the above-mentioned battery insulation layer missing detection method may further include the following step S181:

[0172] Step S181, using the training image set to train the target detection model to be trained, to obtain a trained target detection model; each training image in the training image set has a labeling box for marking the position of the detected target, and during the training process, positive and negative samples are allocated based on the intersection-over-union ratio between the predicted box and the labeled box of the detected target, and the distance from the center point of the labeled box to the predicted box, and the loss is calculated using the allocated positive and negative samples.

[0173] Here, the training image set may include multiple image frames (referred to as training images) obtained by collecting images of the insulation layer installation station of the battery pack during the historical production process. The annotation frame of each training image may be manually or automatically annotated by a machine.

[0174] In some implementations, for each production line, a training image set corresponding to the production line may be used to train the target detection model to be trained, thereby obtaining a trained target detection model for the production line.

[0175] In some embodiments, after a production line is changed (for example, the model of battery products produced on the production line changes, the process of producing battery products on the production line changes, etc.), the training image set corresponding to the updated production line can be re-collected, and the target detection model can be retrained using the re-collected training image set.

[0176] In some implementations, image frames in a training image set may be enhanced and then added to the training image set, thereby expanding the number of training images in the training image set and improving the target detection capability of the trained model.

[0177] In some embodiments, the distribution of positive and negative samples based on the intersection-and-union ratio between the predicted frame and the labeled frame of the detected target, and the distance from the center point of the labeled frame to the predicted frame may include at least one of the following: when the intersection-and-union ratio between the predicted frame and the labeled frame of the detected target is greater than the intersection-and-union ratio threshold, the predicted frame is determined as a positive sample; when the intersection-and-union ratio between the predicted frame and the labeled frame of the detected target is not greater than the intersection-and-union ratio threshold, the positive and negative samples are distributed based on the distance from the center point of the labeled frame to the predicted frame, wherein when the distance from the center point of the labeled frame to the predicted frame is less than the distance threshold, the predicted frame is determined to be a positive sample, and when the distance from the center point of the labeled frame to the predicted frame is less than the distance threshold, the predicted frame is determined to be a negative sample. During implementation, the intersection-and-union ratio threshold and the distance threshold may be set by those skilled in the art according to actual conditions, and the embodiments of the present application do not limit this. For example, the intersection-and-union ratio threshold may be 0.5, 0.6, or 0.4, etc. For another example, the distance threshold may be 0.5 times the side length of the grid where the labeled frame is located. The distance from the center point of the annotation box to the prediction box may be the distance from the center point of the annotation box to the center point of the prediction box, or may be the distance from the center point of the annotation box to the grid where the prediction box is located.

[0178] In the above embodiment, during the training of the target detection model, positive and negative samples are allocated based on the intersection-and-union ratio between the predicted box and the labeled box of the detected target, and the distance from the center point of the labeled box to the predicted box. This can increase the number of allocated positive samples and improve the detection accuracy of the trained target detection model.

[0179] In some embodiments, the insulation layer installation station includes an installation operation area corresponding to the product model of the battery product currently being produced.

[0180] In the above step S111, target detection is performed on the material located at the insulating layer installation station in the image frame to obtain the material detection result corresponding to the image frame, which may include the following step S191:

[0181] Step S191, performing target detection on the material located in the installation operation area in the image frame to obtain the material detection result corresponding to the image frame.

[0182] In the above embodiment, since the installation operation area corresponds to the product model of the battery product currently being produced, target detection of the material located in the installation operation area in the image frame can improve the accuracy of detection and greatly reduce the noise interference generated by target detection, thereby further improving the detection rate of missing insulation layers in battery packs during the battery production process.

[0183] The following describes the application of the battery insulation layer missing detection method provided in the embodiment of the present application in actual scenarios.

[0184] The installation process of the insulation layer on the battery pack (such as mica board, mica sheet, polycarbonate insulation sheet) is complicated. The battery pack, insulation layer, and equipment operating environment are all different in different production lines. Battery packs that fail to install the insulation layer will not only increase the risk of electric vehicle fires, but also the self-inspection of production staff and the random inspection of patrol inspectors. Due to objective reasons, it is impossible for patrol inspectors to complete random inspections of multiple battery packs 24 hours a day. Subjectively, production staff are slack, lazy, and lucky, and it is impossible for them to self-inspect each battery pack. Therefore, it is necessary to perform AI detection on the installation process of the battery pack insulation layer to reduce the number of missing battery pack insulation layers.

[0185] On this basis, an embodiment of the present application provides a battery insulation layer missing detection method, which can be used to detect whether the insulation layer of a battery pack is missing, thereby reducing the problem of operators missing the insulation layer when packaging the battery pack, and reducing the problem of inspectors missing the inspection of battery packs with missing insulation layers; and, by adopting a target detection method, there is no need for target tracking, which can reduce the problem of large amount of calculation and low calculation efficiency caused by the complexity of the target tracking algorithm.

[0186] The present application provides a method for detecting a battery insulation layer missing. Figure 4 As shown, the battery insulation layer missing detection method includes the following steps S201 to S207:

[0187] Step S201, obtaining a wire pulling video stream;

[0188] Here, the wire-pulling video stream may include the target image frame sequence in the above-mentioned embodiment. For example, image frames may be extracted from the wire-pulling video at a specific extraction frequency to obtain the target image frame sequence.

[0189] Step S202, performing target detection on the image frame using a target detection model to obtain an installation state sequence;

[0190] Here, for each image frame in the target image frame sequence, target detection can be performed on the material located at the insulation layer installation station in the image frame using the target detection model to obtain the material detection result corresponding to the image frame, and based on the material detection result corresponding to the image frame, the insulation layer installation status corresponding to the image frame can be determined.

[0191] In some embodiments, a single-stage anchor-free target detection model can be used to perform target detection on image frames. For example, the target detection model can use an improved YOLOv5 model. The target detection model is divided according to the number of stages, and can include a one-stage model or a two-stage model. According to whether a predefined anchor box is required, it can include an anchor-based model and an anchor-free model. Generally speaking, the one-stage model has a faster detection speed than the two-stage model without losing detection accuracy. The anchor-based model has a higher detection accuracy in non-extreme data sets without losing detection speed compared to the anchor-free model. See Figure 2 and Figure 3 , the target detection model in the embodiment of the present application uses the YOLO-v5s model as the basic framework, and uses the aforementioned C5 module to replace the feature extraction module C3 in the YOLO-v5s model. Here, the improved target detection model is recorded as YOLO-C5, which is a one-stage and anchor-based target detection algorithm with the characteristics of high precision and high speed. The training speed of this target detection model is relatively fast. In the case of 300 rounds (epoch), the training time is close to that of most one-stage models (such as RetinaNet model, ATSS model) and two-stage models (such as Faster R-CNN model) in 12 rounds. It can be understood that the C5 module enhances the information flow between multiple modules relative to the C3 module, making the gradient flow richer, with 4 times the channel scaling, and using more parameters to extract features overall, which greatly improves the model performance; and, since the insulating layer target is a large target, when using the original YOLO-v5s, it is easy to have a false detection of multiple boxes on the insulating layer target. Replacing the C3 module with the C5 module significantly improved the situation, reducing the false positive rate from 5.3% to 0.8%, greatly reducing the pressure on the subsequent denoising process.

[0192] When using the YOLO-C5 model to process image frames, it is necessary to extract features through the backbone network (Backbone), use the neck network (Neck) for feature fusion, and use the head network (Head) to restore the real target. The post-processing process of the YOLO-C5 model is similar to the post-processing of the YOLO series, which is dimensional transformation, decoding, category score threshold filtering, confidence and category score product threshold filtering, and restoration to the original image size is not greatly suppressed. After the YOLO-C5 model is inferred, the position and category of the target material in the image frame will be obtained (corresponding to the aforementioned physical detection results). There are many possibilities for the category of the target material, such as battery packs, insulation layers in the form of installation, insulation layers in the form of installation, battery box covers, etc. If the target is not detected, it is recorded as None (corresponding to the aforementioned idle state). None plays two important roles in the subsequent definition of the insulation layer installation event: the start state and the end state.

[0193] During the training of the target detection model, the existing training images can be enhanced and added to the training image set as new training images. The image enhancement process may include but is not limited to at least one of mosaic + random affine transformation, image blur, HSV (hue, saturation, brightness, Hue, Saturation, Value) color space enhancement and random horizontal flipping. Mosaic + random affine transformation can, for example, be used to randomly splice 4 640*640 training images to obtain a 1280*1280 large image during training, and then use random affine transformation to restore it to a 640*640 mixed image, and filter the annotation boxes of the mixed image (i.e., the annotation boxes corresponding to each training image before splicing), remove the annotation boxes with a length or width less than 20 pixels, remove the annotation boxes with an area ratio of less than 0.1 between the enhanced annotation box and the unenhanced annotation box, and retain the annotation boxes with a length-to-width ratio less than 20 and a width-to-length ratio less than 20. After mosaic + random affine transformation, the mixed image is mixed by a combination of distributed sampling, and the annotation boxes are directly pieced together. Image blurring can be implemented, for example, using the Blur class of the Albu library. The kernel of the blurred image is set to (11, 13), and the default value is (3, 7). Because the insulating layer is a large target, the kernel of the blurred image is also set larger to ensure the effectiveness of the blur operation.

[0194] Take the target detection model as the YOLO-C5 model as an example. Figure 5 A schematic diagram of the implementation architecture of a YOLO-C5 model training method provided in an embodiment of the present application. Figure 5As shown, the method includes image enhancement 21, feature extraction 22, feature fusion 23, positive and negative sample allocation 24, detection box encoding 25 and loss calculation 26. Image enhancement 21 can be implemented by a mosaic module Mosc, a random radiation module Rand-Affine, a mixing module MixUp, an image blur module Blur, an HSV color space enhancement and random horizontal flip module; feature extraction 22 can be implemented by the backbone network of the YOLO-C5 model; feature fusion 23 can be implemented by the neck network of the YOLO-C5 model; positive and negative sample allocation 24 can be implemented by a positive and negative sample allocation layer BBox Assigner; detection box encoding 25 can be implemented by BBox Enc; loss calculation 26 can be implemented by a bounding box loss calculation module box_loss, a confidence loss calculation module obj_loss, a classification loss calculation module cls_loss and a total loss calculation module tol_loss.

[0195] The backbone network has a total of 5 layers, including a 6*6 core Conv layer and 4 stage layers. Except for the last Stage layer, the others are composed of a 3*3 core Conv layer plus a C5 module. The last Stage layer adds a spatial pyramid pooling (SPPF) module in addition to a 3*3 core Conv layer plus a C5 module. The SPPF module consists of a Conv layer plus 3 serial 5*5 size two-dimensional maximum pooling (MaxPool2d) layers plus a Conv layer. After the 2nd to 4th Stage layers, a feature map (corresponding to the output feature map in the aforementioned embodiment) will be output respectively to enter the bottleneck network. Taking a 640*640 training image as an example, the size of its output features is (B, 80, 80, 256), (B, 40, 40, 512) and (B, 20, 20, 1024), and the corresponding strides are 8 / 16 / 32 respectively.

[0196] The neck network mainly includes Conv, Upsample, C5 and other modules. Conv is responsible for channel conversion and feature extraction. Upsample is responsible for upsampling and expanding feature maps. Compared with the C3 module, the C5 module makes the gradient flow during feature extraction richer and enhances the information flow between modules. The feature map output by the neck network flows into the head network, and the feature map output by the neck network is input into the three Conv2d layers in the head network that do not share weights. Continuing with the 640*640 training image as an example, the size of the feature map output by the neck network is still (B, 80, 80, 256), (B, 40, 40, 512) and (B, 20, 20, 1024).

[0197] The three Conv2d layers of the head network are responsible for transforming the channels of the feature map, c=(5+n_cls)*3, where n_cls is the number of categories contained in the training image set. There are 15 categories in the training image set, n_cls=15, and the above 5=4+1, where 4 represents the coordinates of the upper left and lower right points of the four boxes, 1 represents the confidence, and 3 represents the preset three anchor boxes. The tensors output by the head network are (B,80,80,3*(4+1+15)), (B,40,40,3*(4+1+15)) and (B,20,20,3*(4+1+15)).

[0198] The positive and negative sample allocation layer BBox Assigner improves the positive and negative sample allocation strategy of YOLOv5 in terms of cross-domain grid strategy, and uses the distance from the center point of the annotation box GT Bbox to the grid to be less than 0.5 times the grid side length to increase the positive samples. The positive and negative sample allocation strategy is divided into the following two core steps:

[0199] 1) Sampling shape rule matching, that is, calculating the width ratio and height ratio of the annotation box GT Bbox and the anchor box Anchor of the current layer. If the maximum value of the width ratio and the height ratio is greater than the set threshold 4, it means that the GT Bbox and the Anchor are not matched enough, and the GT Bbox is temporarily discarded. The predicted position in the grid corresponding to the GT Bbox in the prediction layer is considered to be a negative sample.

[0200] 2) If the maximum value of the width ratio and the height ratio is not greater than the set threshold 4, it means that the GT Bbox and the Anchor match. For the matched GT Bbox, calculate in which grid the GT Bbox falls, and use the distance from the center point of the GT Bbox to the grid to be less than or equal to 0.5 times the grid side length to find the nearest 2 to 4 grids, plus the center point, a total of 3 to 5 grids, all of which are used to predict the GT Bbox (i.e., as positive samples).

[0201] See also Figure 6 and Figure 7 , (a) and (c) are effect diagrams of positive samples generated by the positive-negative sample allocation strategy of YOLOv5, (b) and (d) are effect diagrams of positive samples generated by the improved positive-negative sample allocation strategy of the embodiment of the present application. It can be seen that the number of second positive samples 32 generated by the improved positive-negative sample allocation strategy in the embodiment of the present application is increased compared to the number of first positive samples 31 generated by the positive-negative sample allocation strategy of YOLOv5.

[0202] Compared with the positive and negative sample allocation strategy of YOLOv5, the improved positive and negative sample allocation strategy adopted in the embodiment of the present application generates at least 1.3 times more positive samples, which effectively alleviates the problem of imbalance of positive and negative samples faced by target detection.

[0203] The loss function of YOLO-C5 follows the loss function of the YOLOv3 model. The training mainly includes three aspects of loss: bounding box loss, confidence loss, and classification loss.

[0204] Step S203, sliding window denoising;

[0205] Step S204, removing duplicate adjacent states;

[0206] Step S205, determining whether the battery pack insulation layer is missing;

[0207] If yes, go to step S206, if no, go to step S207.

[0208] Here, a correct battery pack insulation layer installation event can be defined as the following process:

[0209] 1) The workers in the production line are in an idle state. No battery pack, insulation layer in the installation state, or insulation layer in the installation completed state is detected in the production line. This state is recorded as None;

[0210] 2) The battery pack appears in the production line, and this state is recorded as 'battery pack';

[0211] 3) Workers begin to install the insulation layer on the battery pack. The insulation layer is detected in the production line and is in the state of being installed. This state is recorded as "insulation layer being installed".

[0212] 4) The insulation layer is installed. The insulation layer is detected in the production line in the installed state, and this state is recorded as 'insulation layer'.

[0213] 5) The battery packs with installed insulation layers in the production line are transported away, and the workers are idle again. No battery packs, insulation layers in the process of being installed, or insulation layers in the completed state are detected in the production line. This state is also recorded as None.

[0214] If the above process is represented by a state sequence, a correct battery pack insulation layer installation process can be recorded as a target state sequence: [None, 'battery pack', 'insulation layer installation', 'insulation layer', None]. Using the target detection algorithm to extract frames for the battery pack insulation layer installation event, a video frame target sequence (corresponding to the first candidate state sequence in the aforementioned embodiment) is obtained. Since the insulation layer target detection model YOLO-C5 may have a certain false detection rate, the video frame target sequence has noise, for example, see Figure 8, the video frame target sequence 41 may include: [None, None, None, 'battery pack', 'battery pack', 'battery pack', 'insulating layer installing', 'insulating layer installing', 'insulating layer installing', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer installing', 'insulating layer', 'insulating layer', 'insulating layer', None, None, None, None], where the 'insulating layer installing' between 'insulating layer' and 'insulating layer' is noise. Using a sliding window 42 with a default length of 6, the noise can be eliminated by smoothing the video frame target sequence 41 using the state with the most states in the sliding window 42, and the following first state sequence 43 after noise reduction (corresponding to the second candidate state sequence in the aforementioned embodiment) is obtained: [None, None, None, None, None, None, 'battery pack', 'battery pack', 'battery pack', 'insulating layer installing', 'insulating layer installing', 'insulating layer installing', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer', 'insulating layer', None]. The length of the sliding window can be an empirical value, for example, the default length is 6, and the user can also configure an appropriate sliding window length according to the efficiency of production line workers in moving insulating layers, installing insulating layers, and other actions. The first state sequence 43 after noise reduction is deduplicated using adjacent states to obtain a second state sequence 44 after deduplication (corresponding to the installation state sequence in the aforementioned embodiment): [None, 'battery pack', 'insulating layer installing', 'insulating layer', None], which is a correct battery pack insulating layer installation process.

[0215] After a battery pack insulation layer installation event occurs, the production line workers are idle at this time, and the YOLO-C5 model will continue to detect None. These Nones represent the end of the previous installation event and the beginning of the next installation event until the workers continue to be busy. If the workers do not wait for the insulation layer to be installed before transporting the battery pack, the installation state sequence of this installation event is expressed as: [None, 'battery pack', 'insulation layer installing', None], then this is an incorrect battery pack insulation layer installation event.

[0216] In some embodiments, in an actual production environment, there are a large number of battery packs and insulating layers, and one camera is only responsible for detecting the insulating layer installation process of one battery pack. Therefore, a configurable working area (corresponding to the installation operation area in the aforementioned embodiment) is defined, and the battery packs in which the insulating layer is being installed can be detected in the working area to reduce the noise of other battery packs and insulating layers. And only when the installation state sequence detected completely in the working area is it considered to require attention, which effectively reduces the noise caused by complex actions such as intrusion, lifting and moving.

[0217] In some embodiments, the battery pack insulation layer missing event can be confirmed by comparing whether the deduplicated state sequence is consistent with the target state sequence. The event start state and end state are both defined as None. After configuring the work area, if the state sequence obtained after deduplication of the sliding window and adjacent states is [None, 'battery pack' / 'insulation layer being installed' / 'insulation layer', ..., 'battery pack' / 'insulation layer being installed' / 'insulation layer', None], that is, the beginning and the end are both None, which means that the state sequence has constituted an event. If the event is consistent with the target state sequence [None, 'battery pack', 'insulation layer being installed', 'insulation layer', None], it is determined that the battery pack insulation layer is correctly installed (that is, the insulation layer missing event does not occur), otherwise it indicates that the battery pack insulation layer is missing (that is, the insulation layer missing event occurs).

[0218] Step S206, giving an alarm and outputting the missing image frame.

[0219] Step S207, archiving the test results.

[0220] When it is confirmed that the insulation layer of the battery pack is missing, all or part of the image frames corresponding to the target sequence of video frames representing the insulation layer missing event will be alarmed. For example, when the 'insulation layer is installed' state is missing, the video frame corresponding to the 'insulation layer is installed' state will be alarmed. The alarm message can be pushed to at least one of the communication software clusters, such as the fire protection platform, the on-site alarm, the work communication software terminal, etc., to achieve timely processing of the insulation layer missing event, and the image frame sequence reflecting the entire missing process will be saved to the target path for archiving.

[0221] In some embodiments, when the video frame target sequence is determined to be a correctly installed battery pack insulation layer, the image frame sequence corresponding to the video frame target sequence and the corresponding detection results can be saved to the target path for archiving.

[0222] In some embodiments, during the training of the target detection model, the training image set can be obtained by incremental data acquisition. The data of the training image set in this scenario comes from the on-site monitoring camera, and the real-time video stream data of the monitoring camera can be obtained by calling the monitoring platform service. From each battery pack insulation layer installation pull line, the video containing the complete battery pack insulation layer installation process is intercepted, and then the installation state key frame image (i.e., training image) is extracted to form a training image set. After manual annotation, each image generates a text file with the location and category information of the target as an annotation. The category information of the target includes multiple categories such as battery packs corresponding to various battery models, insulation layers in the form of installation, insulation layers in the form of installation, and battery box covers. In the actual factory operation environment, even if the target detection model runs stably, a new pull line will be added, and a new battery pack insulation layer installation process will be added. Limited by the weak generalization ability of the small target detection model, the target detection model needs to be retrained when the pull line and the battery pack insulation layer installation process change, so it is necessary to collect training image data incrementally. When a new installation status key frame image is added, it will be compared with the existing installation status key frame image set (i.e., the existing training image set). If it is not in the existing installation status key frame image set, it will be retained; otherwise, it will be discarded. The retained installation status key frame images will be initially annotated using the target detection model, and then manually adjusted, and finally merged into the existing installation status key frame image set to form a new installation status key frame image set. The new installation status key frame image set will include a new battery pack insulation layer installation process, a new wire pulling scene, and so on. In some embodiments, one image in the installation status key frame image set contains only one target, and the total number of each target is 50.

[0223] In some embodiments, the battery packs produced by the production line are updated according to customer needs, and some frequent changes may be once every two weeks. For example, after the battery packs produced by the production line are changed, the installation process and processing site of the battery pack insulation layer are the same, and the shapes of the battery pack and the insulation layer are different. For another example, after the battery packs produced by the production line are changed, the processing site during the installation process of the battery pack insulation layer is the same, the shapes of the battery pack and the insulation layer are different, and the insulation layer installation process has also changed. Limited by the generalization ability of the current target detection model, it is necessary to re-collect the production line battery pack insulation layer installation images, retrain the model, reconfigure the sliding window length and working area, and redeploy the target detection model. In some embodiments, due to the collaboration of production line workers and business personnel, developers use automated scripts to quickly complete the changes to the target detection model.

[0224] In some embodiments, see Fig. 9 The process of retraining the target detection model includes the following steps S301 to S306:

[0225] Step S301, data analysis;

[0226] Step S302, business confirmation;

[0227] Step S303, determining whether to train a target detection model;

[0228] In some embodiments, the archived detection results in the current production scenario can be analyzed, and it can be determined whether to train the target detection model based on the analysis results. For example, if the missed detection rate of the insulation layer missing event in the archived detection results in the current production scenario exceeds the missed detection rate threshold and / or the false detection rate of the insulation layer missing event exceeds the false detection rate threshold, it indicates that the current target detection model is no longer applicable to the current scenario, and the current target detection model can be retrained.

[0229] In some embodiments, when a production line is changed, an operator can manually confirm whether to train the current object detection model. For example, if the morphology of the battery pack and the insulation layer in the battery products produced by the production line has changed before and after the change, it means that the current object detection model is no longer applicable to the current scenario, and the current object detection model can be retrained.

[0230] Step S304, video data acquisition;

[0231] Step S305, data labeling;

[0232] Step S306, training the target detection model.

[0233] Here, if it is determined that the current target detection model needs to be retrained, video data can be collected from the current production scene, and multiple image frames in the collected video data can be labeled with targets to obtain a training image set. The target detection model is retrained using the training image set, so that a trained target detection model suitable for the current production scene can be obtained.

[0234] The trained target detection model can be deployed or redeployed on a computer device used to perform battery pack insulation layer missing detection, so that target detection can be performed using the trained or retrained target detection model during the battery pack insulation layer missing detection process.

[0235] The target detection model in the embodiment of the present application, on the one hand, improves the core feature extraction module C3 of YOLO-v5s to the C5 module in view of the high false alarm rate on the insulation layer installation data set, and the target detection model implemented greatly reduces the false alarm rate, which reduces the pressure for subsequent denoising; on the other hand, improves the positive and negative sample allocation strategy of YOLO-v5s, and adds a strategy to allocate grids with a distance to the center point of the GT Bbox less than or equal to 0.5 times the grid side length as positive samples, which increases the number of positive samples by at least 1.3 times compared to the positive and negative sample allocation strategy of YOLO-v5s. It effectively improves the accuracy of the insulation layer installation data set with only 750 pictures, greatly reduces the noise such as missed detection and false detection, reduces the pressure for subsequent denoising, and reduces the cost of manual annotation.

[0236] The battery insulation layer missing detection method in the embodiment of the present application uses the target detection YOLO-C5 model to detect the insulation layer installation status in real time, collects various state data of the insulation layer installation process for training, and realizes real-time detection of multiple states such as battery pack, insulation layer installation, and installed insulation layer. There are some unavoidable missed detection and false detection noises in the installation state sequence obtained by processing the image frame sequence using the YOLO-C5 model, and the installation state sequence generated in real time is processed using a sliding window to remove the noise.

[0237] For different production lines, the length and working area of ​​the sliding window are configured according to the workers' actual insulation layer installation speed and area, which greatly reduces the noise generated by target detection. By using the target detection model, there is no need for target tracking, which improves the computational efficiency of the battery insulation layer missing detection method.

[0238] In addition, the battery insulation layer missing detection method in the embodiment of the present application captures the battery pack insulation board missing process, and sends the missing video frame of the missing process to the group and the fire protection platform in the form of an alarm, and synchronously calls the on-site voice alarm player to promptly report and handle the missing illegal operation to reduce quality risks. When there is a major change in the battery pack insulation layer installation product on the production line, the staff will notify the algorithm personnel to retrain the YOLO-C5 target detection model to reduce the problem of the entire battery insulation layer missing detection method failing due to the large amount of noise generated by the target detection model.

[0239] The present application embodiment provides a battery insulation layer missing detection device, such as Fig.10 As shown, the battery insulation layer missing detection device 500 includes a first acquisition module 510, an identification module 520 and a detection module 530, wherein:

[0240] A first acquisition module 510 is used to acquire a target image frame sequence, where the target image frame sequence is obtained by acquiring images of an insulation layer installation station of a battery pack within a target time range;

[0241] The identification module 520 is used to identify the installation state of the insulation layer in each image frame in the target image frame sequence to obtain an installation state sequence; the installation state sequence includes the installation states of the insulation layer corresponding to the multiple image frames respectively;

[0242] The detection module 530 is used to detect the insulation layer installation process within the target time range based on the installation state sequence to determine the detection result of the insulation layer missing event within the target time range.

[0243] In the battery insulation layer missing detection device of the embodiment of the present application, since the insulation layer installation status in each image frame can more completely and directly reflect the insulation layer installation process, based on the installation status sequence, it is possible to more accurately detect whether the insulation layer missing event occurs within the target time range, thereby improving the detection rate of battery pack insulation layer missing during battery production; and, since the detection process does not require target tracking, the hardware computing cost of the detection is reduced and the computing efficiency is improved. In addition, the battery insulation layer missing detection device of the embodiment of the present application can automatically detect whether the insulation layer missing event occurs within the target time range, thereby reducing the workload of manual visual inspection and reducing labor costs.

[0244] In some embodiments, the recognition module is also used to: for each image frame in the target image frame sequence, perform target detection on the material located at the insulation layer installation station in the image frame to obtain the material detection result corresponding to the image frame, and determine the insulation layer installation status corresponding to the image frame based on the material detection result corresponding to the image frame; determine the installation status sequence based on the insulation layer installation status corresponding to each image frame.

[0245] In some embodiments, the insulating layer installation station is used to install an insulating layer on a first material, and the insulating layer and the first material are used to be assembled with a second material to obtain a target battery product. The first material includes a battery pack or a battery box cover, and the second material includes one of the battery pack and the battery box cover that is different from the first material.

[0246] In some embodiments, the identification module is also used to: when the material detection result corresponding to the image frame includes a first detection result, determine that the installation status of the insulation layer corresponding to the image frame is that the first material has been put online, and the first detection result indicates that the first material has been detected.

[0247] In some embodiments, the recognition module is also used to: when the target detection result corresponding to the image frame includes a second detection result, determine that the insulation layer installation state corresponding to the image frame is the insulation layer being installed, and the second detection result represents that the insulation layer is detected in the installation state.

[0248] In some embodiments, the recognition module is also used to: when the target detection result corresponding to the image frame includes a third detection result, determine that the installation state of the insulation layer corresponding to the image frame is that the insulation layer has been installed, and the third detection result indicates that the insulation layer is detected in a completed installation state.

[0249] In some embodiments, the identification module is also used to: when the target detection result corresponding to the image frame includes a fourth detection result, determine that the installation state of the insulating layer corresponding to the image frame is an idle state, and the fourth detection result characterizes that the first material is not detected, the insulating layer is in an installation state, and the insulating layer is in an installation completed state.

[0250] In some embodiments, the recognition module is also used to: determine a first candidate state sequence based on the insulation layer installation state corresponding to each image frame; perform noise reduction processing on the first candidate state sequence based on a sliding window of a target length to obtain a second candidate state sequence; and perform deduplication processing on adjacent insulation layer installation states in the second candidate state sequence to obtain an installation state sequence.

[0251] In some embodiments, the battery insulation layer missing installation detection device also includes: a first determination module, used to determine the operation time of a single installation of the insulation layer in the production process corresponding to the product model of the target battery product based on the product model of the target battery product; a second determination module, used to determine the target length based on the acquisition interval and operation time between adjacent image frames in the target image frame sequence.

[0252] In some embodiments, the detection module is also used to: detect a candidate state subsequence from the installation state sequence, the candidate state subsequence takes the idle state as the starting state and the idle state as the ending state, and the starting state and the ending state include at least one of the following states: the first material has been put online, the insulation layer is being installed, and the insulation layer has been installed; when the candidate state subsequence is inconsistent with the target state sequence, it is determined that an insulation layer missing event occurs within the target time range.

[0253] In some embodiments, the recognition module is also used to: utilize the backbone network in the trained target detection model to extract features from the image frame to obtain multi-size first image features; utilize the neck network in the target detection model to perform multi-scale feature fusion on the first image features to obtain second image features; utilize the head network in the target detection model to predict the position and category of the material located at the insulation layer installation station in the image frame based on the second image features to obtain the material detection results corresponding to the image frame.

[0254] In some embodiments, the backbone network includes N layers of cascaded feature extraction modules, where N is a natural number greater than 2; the recognition module is also used to: perform convolution processing on the image frame to obtain the first input feature map; use the i-th layer feature extraction module to perform the following operations: perform convolution processing on the i-th input feature map to obtain the i-th first intermediate feature map; perform at least one residual connection processing on the i-th first intermediate feature map and then perform convolution processing to obtain the i-th second intermediate feature map; perform at least one residual connection processing on the i-th second intermediate feature map and then perform convolution processing to obtain the i-th third intermediate feature map; perform at least one residual connection operation on the third intermediate feature map to obtain the i-th fourth intermediate feature map; the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map and the i-th fourth intermediate feature map are connected by channel and then convolved to obtain the i-th output feature map; wherein, i is a natural number not exceeding N; the N-th output feature map is pooled to obtain the N-th output feature map after pooling; the first image feature includes the j-th output feature map and the N-th output feature map after pooling, j is a natural number greater than 1 and less than N; the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map, the i-th fourth intermediate feature map and the i-th output feature map have the same size.

[0255] In some embodiments, the battery insulation layer missing detection device also includes: a training module, which is used to use a training image set to train a target detection model to be trained to obtain a trained target detection model; each training image in the training image set has a labeling box for marking the position of the detected target. During the training process, positive and negative samples are allocated based on the intersection and union ratio between the predicted box and the labeled box of the detected target, and the distance from the center point of the labeled box to the predicted box, and the loss is calculated using the allocated positive and negative samples.

[0256] In some embodiments, the insulating layer installation station includes an installation operation area corresponding to the product model of the battery product currently being produced; the identification module is also used to: perform target detection on the material located in the installation operation area in the image frame to obtain the material detection result corresponding to the image frame.

[0257] 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 embodiment of the present application 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 application, please refer to the description of the method embodiment of the present application for understanding.

[0258] It should be noted that in the embodiments of the present application, 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 embodiment of the present application can be essentially or partly reflected in the form of a software product that contributes to the relevant technology. The software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods of each embodiment of the present application. 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 application are not limited to any specific hardware, software or firmware, or any combination of hardware, software, and firmware.

[0259] An embodiment of the present application 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-mentioned battery insulation layer missing detection method are implemented.

[0260] The embodiment of the present application 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-mentioned battery insulation layer missing detection method are implemented. The computer-readable storage medium can be transient or non-transient.

[0261] An embodiment of the present application provides a computer program, including a computer-readable code. When the computer-readable code is run in a computer device, a processor in the computer device executes some or all of the steps in the above method.

[0262] The embodiment of the present application provides 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.

[0263] An embodiment of the present application provides a computer program product, including a computer program or instructions. When the computer program or instructions are executed by a processor, some or all of the steps in the above-mentioned battery insulation layer missing detection method are implemented.

[0264] 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 of this application, please refer to the description of the method embodiment of this application for understanding.

[0265] It should be noted that Fig.11 A schematic diagram of a hardware entity of a computer device in an embodiment of the present application is shown in FIG. Fig.11 As shown, the hardware entity of the computer device 600 includes: a processor 601, a communication interface 602 and a memory 603, wherein: the processor 601 generally controls the overall operation of the computer device 600. The communication interface 602 can enable the computer device to communicate with other terminals or servers through a network. The memory 603 is configured to store instructions and applications executable by the processor 601, and can also cache data to be processed or processed by the processor 601 and each module in the computer device 600 (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 transmission can be performed between the processor 601, the communication interface 602 and the memory 603 through a bus 604.

[0266] 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 application. 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 various embodiments of the present application, the size of the serial number of each step / process mentioned above does not mean the order of execution, and 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 embodiment of the present application. The serial numbers of the embodiments of the present application mentioned above are for description only and do not represent the advantages and disadvantages of the embodiments.

[0267] 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.

[0268] In the several embodiments provided in the present application, 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 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 devices or units can be electrical, mechanical or other forms.

[0269] 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 embodiments of the present application may be all 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.

[0270] 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.

[0271] Alternatively, if the above-mentioned integrated unit of the present application 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 application can essentially or in other words, 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 a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods of each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0272] The above are only exemplary embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A method for detecting a battery insulation layer missing, characterized in that: include: Acquire a target image frame sequence, wherein the target image frame sequence is obtained by performing image acquisition on an insulation layer installation station of a battery pack within a target time range; wherein the insulation layer installation station is used to install an insulation layer on a first material; Identifying the installation state of the insulation layer in each image frame in the target image frame sequence to obtain an installation state sequence; the installation state sequence includes the installation states of the insulation layer corresponding to a plurality of the image frames respectively; Based on the installation state sequence, the insulation layer installation process within the target time range is detected to determine the detection result of the insulation layer missing event within the target time range; The step of identifying the installation state of the insulating layer in each image frame in the target image frame sequence to obtain an installation state sequence includes: For each image frame in the target image frame sequence, target detection is performed on the material located at the insulating layer installation station in the image frame to obtain the material detection result corresponding to the image frame, and based on the material detection result corresponding to the image frame, the insulating layer installation state corresponding to the image frame is determined; based on the insulating layer installation state corresponding to each image frame, the installation state sequence is determined; wherein the material detection result includes at least one of the following: the first material is detected, the insulating layer in the state of being installed is detected, the insulating layer in the state of being installed is detected, and the first material, the insulating layer in the state of being installed, and the insulating layer in the state of being installed are not detected; The detecting the insulation layer installation process within the target time range based on the installation state sequence includes: Detecting a candidate state subsequence from the installation state sequence, wherein the candidate state subsequence takes an idle state as a starting state and takes the idle state as an ending state, and at least one of the following states is included between the starting state and the ending state: the first material is online, the insulation layer is being installed, and the insulation layer is installed; When the candidate state subsequence is inconsistent with the target state sequence, it is determined that an insulation layer missing event occurs within the target time range; wherein, the target state sequence represents an insulation layer installation state sequence in which the insulation layer has been correctly installed, and the target state sequence includes the following insulation layer installation states in sequence: the idle state, the first material has been put online, the insulation layer is being installed, the insulation layer has been installed, and the idle state.

2. The battery insulation layer missing detection method according to claim 1, characterized in that: The insulating layer and the first material are used to be assembled with a second material to obtain a target battery product, the first material includes a battery pack or a battery box cover, and the second material includes one of the battery pack and the battery box cover that is different from the first material.

3. The battery insulation layer missing detection method according to claim 2, characterized in that: The determining, based on the material detection result corresponding to the image frame, the installation state of the insulation layer corresponding to the image frame comprises: In the case where the material detection result corresponding to the image frame includes a first detection result, it is determined that the installation state of the insulation layer corresponding to the image frame is that the first material has been put online; and the first detection result indicates that the first material has been detected.

4. The battery insulation layer missing detection method according to claim 2, characterized in that: The determining, based on the material detection result corresponding to the image frame, the installation state of the insulation layer corresponding to the image frame comprises: When the material detection result corresponding to the image frame includes a second detection result, it is determined that the insulation layer installation state corresponding to the image frame is the insulation layer being installed; the second detection result indicates that the insulation layer is detected to be in the installation state.

5. The battery insulation layer missing detection method according to claim 2, characterized in that: The determining, based on the material detection result corresponding to the image frame, the installation state of the insulation layer corresponding to the image frame comprises: When the material detection result corresponding to the image frame includes a third detection result, it is determined that the installation state of the insulation layer corresponding to the image frame is that the insulation layer has been installed; the third detection result indicates that the insulation layer is detected to be in an installed state.

6. The method for detecting missing battery insulation layer according to claim 2, characterized in that: The determining, based on the material detection result corresponding to the image frame, the installation state of the insulation layer corresponding to the image frame comprises: When the material detection result corresponding to the image frame includes a fourth detection result, it is determined that the installation state of the insulating layer corresponding to the image frame is an idle state; the fourth detection result indicates that the first material, the insulating layer in the installation state, and the insulating layer in the installation completed state are not detected.

7. The method for detecting missing battery insulation layer according to claim 2, characterized in that: The step of determining the installation state sequence based on the installation state of the insulation layer corresponding to each of the image frames comprises: Determining a first candidate state sequence based on the insulating layer installation state corresponding to each of the image frames; Based on a sliding window of a target length, performing noise reduction processing on the first candidate state sequence to obtain a second candidate state sequence; Adjacent insulating layer installation states in the second candidate state sequence are deduplicated to obtain the installation state sequence.

8. The battery insulation layer missing detection method according to claim 7, characterized in that: The battery insulation layer missing detection method further comprises: Based on the product model of the target battery product, determining the operation time of a single installation of an insulating layer in a production process corresponding to the product model; The target length is determined based on a collection interval between adjacent image frames in the target image frame sequence and the operation time.

9. The method for detecting missing battery insulation layer according to claim 1, characterized in that: The performing target detection on the material located at the insulating layer installation station in the image frame to obtain the material detection result corresponding to the image frame includes: Using a backbone network in a trained object detection model, extracting features from the image frame to obtain first image features of multiple sizes; Using the neck network in the target detection model, multi-scale feature fusion is performed on the first image feature to obtain a second image feature; The head network in the target detection model is used to predict the position and category of the material located at the insulation layer installation station in the image frame based on the second image feature to obtain a material detection result corresponding to the image frame.

10. The method for detecting missing battery insulation layer according to claim 9, characterized in that: The backbone network includes N layers of cascaded feature extraction modules, where N is a natural number greater than 2; The method of using a backbone network in a trained target detection model to extract features from the image frame to obtain first image features of multiple sizes includes: Performing convolution processing on the image frame to obtain a first input feature map; The feature extraction module of the i-th layer is used to perform the following operations: convolution processing is performed on the i-th input feature map to obtain the i-th first intermediate feature map; the i-th first intermediate feature map is subjected to at least one residual connection processing and then convolution processing to obtain the i-th second intermediate feature map; the i-th second intermediate feature map is subjected to at least one residual connection processing and then convolution processing to obtain the i-th third intermediate feature map; the third intermediate feature map is subjected to at least one residual connection operation to obtain the i-th fourth intermediate feature map; the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map and the i-th fourth intermediate feature map are connected according to channels and then convolution processing is performed to obtain the i-th output feature map; wherein i is a natural number not exceeding N; The Nth output feature map is pooled to obtain the Nth output feature map after pooling; the first image feature includes the jth output feature map and the Nth output feature map after pooling, where j is a natural number greater than 1 and less than N; the i-th first intermediate feature map, the i-th second intermediate feature map, the i-th third intermediate feature map, the i-th fourth intermediate feature map and the i-th output feature map have the same size.

11. The method for detecting battery insulation layer missing installation according to claim 9, characterized in that: The battery insulation layer missing detection method further comprises: A target detection model to be trained is trained using a training image set to obtain the trained target detection model; each training image in the training image set has a labeling box for marking the position of a detected target, and during the training process, positive and negative samples are allocated based on the intersection-over-union ratio between a predicted box of the detected target and the labeling box, and the distance from the center point of the labeling box to the predicted box, and the loss is calculated using the allocated positive and negative samples.

12. The method for detecting missing battery insulation layer according to any one of claims 1 to 11, characterized in that: The insulating layer installation station includes an installation operation area corresponding to the product model of the battery product currently being produced; The performing target detection on the material located at the insulating layer installation station in the image frame to obtain the material detection result corresponding to the image frame includes: Perform target detection on the material located in the installation operation area in the image frame to obtain a material detection result corresponding to the image frame.

13. A battery insulation layer missing detection device, characterized in that: include: A first acquisition module is used to acquire a target image frame sequence, wherein the target image frame sequence is obtained by performing image acquisition on an insulation layer installation station of a battery pack within a target time range; wherein the insulation layer installation station is used to install an insulation layer on a first material; an identification module, used for identifying the installation state of the insulation layer in each image frame in the target image frame sequence to obtain an installation state sequence; the installation state sequence includes the installation states of the insulation layer corresponding to a plurality of the image frames respectively; A detection module, configured to detect the insulation layer installation process within the target time range based on the installation state sequence, so as to determine the detection result of the insulation layer missing installation event within the target time range; The identification module is specifically used for: For each image frame in the target image frame sequence, target detection is performed on the material located at the insulating layer installation station in the image frame to obtain the material detection result corresponding to the image frame, and based on the material detection result corresponding to the image frame, the insulating layer installation state corresponding to the image frame is determined; based on the insulating layer installation state corresponding to each image frame, the installation state sequence is determined; wherein the material detection result includes at least one of the following: the first material is detected, the insulating layer in the state of being installed is detected, the insulating layer in the state of being installed is detected, and the first material, the insulating layer in the state of being installed, and the insulating layer in the state of being installed are not detected; The detection module is specifically used for: Detecting a candidate state subsequence from the installation state sequence, wherein the candidate state subsequence takes an idle state as a starting state and takes the idle state as an ending state, and at least one of the following states is included between the starting state and the ending state: the first material is online, the insulation layer is being installed, and the insulation layer is installed; When the candidate state subsequence is inconsistent with the target state sequence, it is determined that an insulation layer missing event occurs within the target time range; wherein, the target state sequence represents an insulation layer installation state sequence in which the insulation layer has been correctly installed, and the target state sequence includes the following insulation layer installation states in sequence: the idle state, the first material has been put online, the insulation layer is being installed, the insulation layer has been installed, and the idle state.

14. 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 of the battery insulation layer missing detection method according to any one of claims 1 to 12 are implemented.

15. 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 of the battery insulation layer missing detection method according to any one of claims 1 to 12 are implemented.

16. 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 battery insulation layer missing detection method according to any one of claims 1 to 12 are implemented.

Images