Vehicle detection and tracking method and device
By extracting and fusion of vehicle target detection and tracking technology in highway environment, and combining spatial attention mechanism and re-identification branches, poor tracking effect caused by occlusion and environmental complexity of vehicle targets in highway environment is solved. The problem is that high-precision and real-time vehicle target detection and tracking are achieved.
Patent Information
- Application Number
- CN202210393312.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-15
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2042-04-15
AI Technical Summary
The existing vehicle target detection and tracking technology has poor tracking results in highway environments due to factors such as vehicle shading and environmental complexity, and the processing speed is difficult to meet the real-time monitoring requirements.
A vehicle detection and tracking method is adopted. By extracting and fusion of surveillance videos, inputting a pre-trained object detection and tracking model for vehicle detection and tracking, combining the spatial attention mechanism and re-identification branches, the model's vehicle targets are improved. Detection accuracy and tracking stability.
Effectively reduce vehicle target tracking drift and loss caused by target occlusion, loss, similarity, transformation and other factors, improve tracking accuracy and meet real-time requirements.
Smart Images

Figure CN114897933B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of intelligent transportation, and in particular to a vehicle detection and tracking method and device suitable for highway monitoring videos. Background Art
[0002] At present, vehicle target detection and tracking technology is widely used in intelligent driving, intelligent monitoring and other fields. In the highway environment, the application of vehicle detection and tracking technology and automatic warning can effectively reduce highway vehicle accidents and improve highway operation safety. However, since this field involves many disciplines and the research objects are complex, even after years of development, there are still many problems to be solved.
[0003] Especially when it is actually applied in a highway environment, vehicle targets are difficult to track due to factors such as frequent occlusions between vehicle targets and complex and changeable environments. This type of problem has always been a difficult point in research. Many existing technologies and solutions are not perfect enough and are far from being practical.
[0004] At present, most of the vehicle target tracking methods using deep learning are based on the detection tracking paradigm. The idea of the detection-based tracking method is to first detect the vehicle target in the current frame, and then match the detected vehicle target with the previous vehicle trajectory. The characteristics of this type of method are that the target tracking effect is heavily dependent on the detection effect, and due to the lack of shared computing between the detection task and the tracking task, the model processing speed is slow, making it difficult to meet the requirements of real-time monitoring. Summary of the invention
[0005] In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a vehicle detection and tracking method and device to solve at least one defect in the prior art.
[0006] To achieve the above objectives and other related objectives, the present invention provides a vehicle detection and tracking method, comprising:
[0007] Extract the surveillance video to obtain the current frame image, the previous frame image of the current frame, and the heat map of the previous frame image;
[0008] Extracting features from the current frame image, the previous frame image, and the heat map of the previous frame image respectively, and fusing the extracted features to obtain a fused feature map;
[0009] The fused feature map is input into a pre-trained target detection and tracking model to achieve detection and tracking of the vehicle.
[0010] Optionally, the target detection and tracking model is a model based on CenterTrack, and the backbone network of the target detection and tracking model is DLA-34; the target detection and tracking model includes a first output branch, a second output branch, and a third output branch; the backbone network is used to extract features from the fused feature map, and the feature map extracted by the backbone network is output through the first output branch, the second output branch, and the third output branch, respectively, as a target heat map feature map, a target bounding box feature map, and a vehicle target re-identification feature map.
[0011] Optionally, after the backbone network extracts features from the fused feature map, the method further includes:
[0012] The extracted feature map is processed based on the spatial attention mechanism to obtain the target feature map;
[0013] Optionally, respectively extracting features from the current frame image, the previous frame image, and the heat map of the previous frame image, and fusing the extracted features to obtain a fused feature map, includes:
[0014] Performing convolution, batch normalization and ReLU on the current frame image to obtain a first feature map;
[0015] Performing convolution, batch normalization and ReLU on the previous frame image to obtain a second feature map;
[0016] According to the distribution of vehicle targets in the previous frame image, a heat map of the previous frame image is obtained, and convolution, batch normalization and ReLU are performed on the heat map to obtain a third feature map;
[0017] The first feature map, the second feature map, and the third feature map are connected according to the channel dimension to obtain a fused feature map.
[0018] Optionally, the processing the fused feature map based on the spatial attention mechanism to obtain a target feature map includes:
[0019] Performing feature extraction on the fused feature map to obtain a feature map A;
[0020] Performing maximum pooling and average pooling on the feature map A respectively to obtain feature map A1 and feature map A2, and connecting the feature map A1 and feature map A2 by channel to obtain feature map B;
[0021] Perform convolution and Sigmoid function processing on the feature map B to obtain a feature map C;
[0022] Multiply each channel of feature map A and feature map C bit by bit to obtain the target feature map.
[0023] Optionally, the re-identification branch is a linear connection layer, the input dimension of the re-identification branch is the total number of non-repeated vehicle targets appearing in the training set, and the output dimension is 128.
[0024] Optionally, the training method of the target detection and tracking model includes:
[0025] The current frame image, the previous frame image and the heat map of the previous frame image are used as inputs, and the target heat map, target regression result, offset regression result and target re-identification result are used as outputs to train the initial network and obtain the initial target detection and tracking model;
[0026] The loss value of the target detection and tracking model is calculated based on the predicted output and true output of the target detection and tracking model. The Adam optimizer is used to train several epochs on the training set until the stopping condition is met. The final target detection and tracking model and the weight of the target detection and tracking model are obtained and saved.
[0027] Optionally, the re-identification branch of the target detection and tracking model outputs a feature map I, performs a mapping operation on the feature map I, calculates a vector tracking_id, and the vector tracking_id is passed through a linear classifier to obtain an nID-dimensional vector tracking_output, where nID is the number of different targets in the training set;
[0028] Calculate the nID-dimensional vector target_id based on the true label of the training set;
[0029] Use cross entropy loss function Loss ReID Calculate the training loss of the re-ID branch;
[0030] Loss ReID =CrossEntropyLoss(tracking_id,target_id)
[0031] CrossEntropyLoss(x,y)={l1,...,l N} T
[0032]
[0033] Among them, x is the input, y is the target, ω is the weight, and N is the set minimum batch size.
[0034] To achieve the above objectives and other related objectives, the present invention provides a vehicle detection and tracking device, comprising:
[0035] An image extraction module is used to extract the surveillance video to obtain the current frame image, the previous frame image of the current frame, and the heat map of the previous frame image;
[0036] A feature fusion module, used to extract features from the current frame image, the previous frame image, and the heat map of the previous frame image, and fuse the extracted features to obtain a fused feature map;
[0037] The target detection module is used to input the fused feature map into a pre-trained target detection and tracking model to achieve detection and tracking of the vehicle.
[0038] As described above, a vehicle detection and tracking method and device of the present invention has the following beneficial effects:
[0039] A vehicle detection and tracking method of the present invention includes: extracting monitoring video to obtain a current frame image, a previous frame image of the current frame, and a heat map of the previous frame image; extracting features from the current frame image, the previous frame image, and the heat map of the previous frame image, respectively, and fusing the extracted features to obtain a fused feature map; inputting the fused feature map into a pre-trained target detection and tracking model to achieve vehicle detection and tracking. The vehicle target tracking method proposed by the present invention, which is suitable for the characteristics of the highway off-site environment, can effectively reduce vehicle target tracking drift and tracking loss caused by factors such as target occlusion, target loss, target similarity, and target transformation, improve tracking accuracy, and meet real-time requirements. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 A flow chart of a vehicle detection and tracking method according to an embodiment of the present invention;
[0041] Figure 2 This is a flow chart of a vehicle detection and tracking method according to another embodiment of the present invention;
[0042] Figure 3 A flow chart of a method for processing the fused feature map based on a spatial attention mechanism according to an embodiment of the present invention;
[0043] Figure 4 A flowchart of obtaining a re-identification output branch according to an embodiment of the present invention;
[0044] Figure 5 A flow chart of obtaining a fusion feature map according to an embodiment of the present invention;
[0045] Figure 6 The figure is a schematic structural diagram of a vehicle detection and tracking device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0046] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.
[0047] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.
[0048] like Figure 1 , 2 As shown, the embodiment of the present application provides a vehicle detection and tracking method, including:
[0049] S100 extracts the surveillance video to obtain a current frame image, a previous frame image of the current frame, and a heat map of the previous frame image;
[0050] S200 extracts features from the current frame image, the previous frame image, and the heat map of the previous frame image respectively, and fuses the extracted features to obtain a fused feature map;
[0051] S300 inputs the fused feature map into a pre-trained target detection and tracking model to achieve detection and tracking of the vehicle.
[0052] In one embodiment, the target detection and tracking model is a model based on CenterTrack, and the backbone network of the target detection and tracking model is DLA-34; the target detection and tracking model includes a first output branch, a second output branch, and a third output branch, the first output branch is used to output a target heat map feature map, the second output branch is used to output a target bounding box feature map, and the third output branch is used to output a target re-identification feature map; Figure 1 In the post-processing, the feature maps output by the three branches are encoded to obtain the category, bounding box coordinates and tracking ID information of the vehicle target. The present invention improves the recognition ability of the target detection and tracking model for different vehicle targets by adding a target re-identification branch.
[0053] In one embodiment, after the backbone network extracts features from the fused feature map, the method further includes:
[0054] The extracted feature map is processed based on the spatial attention mechanism to obtain a target feature map. The first output branch, the second output branch, and the third output branch process the target feature map, and then post-processing is performed to achieve vehicle detection and tracking.
[0055] By designing spatial attention, the trained target detection and tracking model can focus on areas where vehicle targets are more likely to appear, thereby improving the detection effect of the model's vehicle targets.
[0056] Specifically, the feature map extracted by the backbone network is processed based on the spatial attention mechanism to obtain the target feature map, including:
[0057] Perform feature extraction on the feature map extracted by the backbone network to obtain feature map A;
[0058] Performing maximum pooling and average pooling on the feature map A respectively to obtain feature map A1 and feature map A2, and connecting the feature map A1 and feature map A2 by channel to obtain feature map B;
[0059] Perform convolution and Sigmoid function processing on the feature map B to obtain a feature map C;
[0060] Multiply each channel of feature map A and feature map C bit by bit to obtain the target feature map.
[0061] More specifically, Figure 3 As shown, the feature map extracted by the backbone network is subjected to feature extraction to obtain feature map A. Feature map A is subjected to 3×3 maximum pooling to obtain Feature map A1, average pooling is performed on the feature map to obtain Feature map A2;
[0062] Connect feature map A1 and feature map A2 according to the channel dimension to obtain Feature map B, perform 7×7 convolution on feature map B, and use Sigmoid function to process it, and we get Feature map C;
[0063] Multiply the feature map A by the channel features of the feature map C bit by bit to obtain The feature map is taken and output, and the target feature map is obtained.
[0064] In one embodiment, the re-identification branch is a linear connection layer, the input dimension of the re-identification branch is the total number of non-repeated vehicle targets appearing in the training set, and the output dimension is 128.
[0065] like Figure 4As shown in the figure, the re-identification branch of the target detection and tracking model outputs a feature map I, maps the feature map I, calculates the vector tracking_id, and the vector tracking_id is passed through a linear classifier to obtain an nID-dimensional vector tracking_output, where nID is the number of different targets in the training set.
[0066] Calculate the nID-dimensional vector target_id based on the true label of the training set;
[0067] Use cross entropy loss function Loss ReID Calculate the training loss of the re-ID branch;
[0068] Loss ReID =CrossEntropyLoss(tracking_id,target_id)
[0069] CrossEntropyLoss(x,y)={l1,...,l N} T
[0070]
[0071] Among them, x is the input, y is the target, ω is the weight, and N is the set minimum batch size.
[0072] In one embodiment, the extracting features of the current frame image, the previous frame image, and the heat map of the previous frame image respectively, and fusing the extracted features to obtain a fused feature map, includes:
[0073] Performing convolution, batch normalization and ReLU on the current frame image to obtain a first feature map;
[0074] Performing convolution, batch normalization and ReLU on the previous frame image to obtain a second feature map;
[0075] According to the distribution of vehicle targets in the previous frame image, a heat map of the previous frame image is obtained, and convolution, batch normalization and ReLU are performed on the heat map to obtain a third feature map;
[0076] The first feature map, the second feature map, and the third feature map are connected according to the channel dimension to obtain a fused feature map.
[0077] Specifically, Figure 5 As shown, the 3×H×W image of the current frame is subjected to 7×7 convolution, batch normalization and ReLU to obtain the 16×H×W first feature map;
[0078] Determine whether the previous frame image exists. If so, perform 7×7 convolution, batch normalization and ReLU on the previous frame image to obtain a 16×H×W second feature map. According to the distribution of vehicle targets in the previous frame image, obtain the heat map of the previous frame image. Perform 7×7 convolution, batch normalization and ReLU on the heat map to obtain a 16×H×W third feature map. Connect the first feature map, the second feature map and the third feature map according to the channel dimension to obtain a 48×H×W feature map and output it.
[0079] If the previous frame image exists, a 0-valued tensor 32×H×W is generated, and the first feature map is concatenated with this tensor according to the channel dimension to obtain a 48×H×W feature map and output it.
[0080] In one embodiment, the training method of the target detection and tracking model includes:
[0081] The current frame image, the previous frame image and the heat map of the previous frame image are used as inputs, and the target heat map, target regression result, offset regression result and target re-identification result are used as outputs to train the initial network and obtain the initial target detection and tracking model;
[0082] The loss value of the target detection and tracking model is calculated based on the predicted output and true output of the target detection and tracking model. The Adam optimizer is used to train several epochs on the training set until the stopping condition is met. The final target detection and tracking model and the weight of the target detection and tracking model are obtained and saved.
[0083] After completing the training of the target detection and tracking model, use the designed target detection and tracking model to load the weights of the target detection and tracking model; perform frame-by-frame vehicle target detection and tracking on the input monitoring video stream data.
[0084] The present invention makes full use of the features extracted by the backbone network by jointly training the target detection and tracking model to accelerate the vehicle target tracking speed; by fusing multiple input features, the feature map information from multiple inputs is better integrated, thereby improving the detection and tracking accuracy; by adding spatial attention, the target detection and tracking model focuses on the area where the vehicle target has a higher probability of appearing, thereby improving the detection accuracy of the vehicle target; by adding a vehicle re-identification output branch to the target detection and tracking model, the target detection and tracking model is improved in terms of its ability to identify different vehicle targets, thereby improving the tracking accuracy of the target detection and tracking model. Afterwards, the designed model is trained on the prepared highway vehicle tracking data set until the model converges and the trained model weights are saved; based on the tracking model and the trained weights, vehicle target detection and tracking in the monitoring video under the highway can be achieved. The method can effectively reduce the vehicle target tracking drift and tracking loss caused by factors such as target occlusion, target loss, target similarity, and target transformation, improve the tracking accuracy, and meet the real-time requirements.
[0085] like Figure 6 As shown, the embodiment of the present application provides a vehicle detection and tracking device, comprising:
[0086] An image extraction module is used to extract the surveillance video to obtain the current frame image, the previous frame image of the current frame, and the heat map of the previous frame image;
[0087] A feature fusion module, used to extract features from the current frame image, the previous frame image, and the heat map of the previous frame image, and fuse the extracted features to obtain a fused feature map;
[0088] The target detection module is used to input the fused feature map into a pre-trained target detection and tracking model to achieve detection and tracking of the vehicle.
[0089] The specific implementation methods of the above-mentioned device and detection method are substantially the same and will not be described in detail here.
[0090] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.
[0091] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0092] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0093] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic. For example, the division of the modules or 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 integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0094] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, 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 solution of this embodiment.
[0095] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0096] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory ((RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.
[0097] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.
Claims
1. A vehicle detection and tracking method, characterized in that: include: Extract the surveillance video to obtain the current frame image, the previous frame image of the current frame, and the heat map of the previous frame image; Extracting features from the current frame image, the previous frame image, and the heat map of the previous frame image respectively, and fusing the extracted features to obtain a fused feature map; Inputting the fused feature map into a pre-trained target detection and tracking model to achieve vehicle detection and tracking; The target detection and tracking model is a model based on CenterTrack, and the backbone network of the target detection and tracking model is DLA-34; the target detection and tracking model includes a first output branch, a second output branch, and a third output branch; the backbone network is used to extract features from the fused feature map, and the feature map extracted by the backbone network is output through the first output branch, the second output branch, and the third output branch to output a target heat map feature map, a target bounding box feature map, and a vehicle target re-identification feature map respectively; After the backbone network extracts features from the fused feature map, the method further includes: processing the extracted feature map based on a spatial attention mechanism to obtain a target feature map; The processing of the fused feature map based on the spatial attention mechanism to obtain a target feature map includes: Performing feature extraction on the fused feature map to obtain a feature map A; Performing maximum pooling and average pooling on the feature map A respectively to obtain feature map A1 and feature map A2, and connecting the feature map A1 and feature map A2 by channel to obtain feature map B; Perform convolution and Sigmoid function processing on the feature map B to obtain a feature map C; Multiply each channel of feature map A and feature map C bit by bit to obtain the target feature map.
2. The vehicle detection and tracking method according to claim 1, characterized in that: The extracting features of the current frame image, the previous frame image, and the heat map of the previous frame image respectively, and fusing the extracted features to obtain a fused feature map, comprises: Performing convolution, batch normalization and ReLU on the current frame image to obtain a first feature map; Performing convolution, batch normalization and ReLU on the previous frame image to obtain a second feature map; According to the distribution of vehicle targets in the previous frame image, a heat map of the previous frame image is obtained, and convolution, batch normalization and ReLU are performed on the heat map to obtain a third feature map; The first feature map, the second feature map, and the third feature map are connected according to the channel dimension to obtain a fused feature map.
3. The vehicle detection and tracking method according to claim 1, characterized in that: The target detection and tracking model also includes a re-identification branch, which is a linear connection layer. The input dimension of the re-identification branch is the total number of non-repeated vehicle targets appearing in the training set, and the output dimension is 128.
4. The vehicle detection and tracking method according to claim 1, characterized in that: The training method of the target detection and tracking model includes: The current frame image, the previous frame image and the heat map of the previous frame image are used as inputs, and the target heat map, target regression result, offset regression result and target re-identification result are used as outputs to train the initial network and obtain the initial target detection and tracking model; The loss value of the target detection and tracking model is calculated based on the predicted output and true output of the target detection and tracking model. The Adam optimizer is used to train several epochs on the training set until the stopping condition is met. The final target detection and tracking model and the weight of the target detection and tracking model are obtained and saved.
5. The vehicle detection and tracking method according to claim 4, characterized in that: The re-identification branch of the target detection and tracking model outputs a feature map I, maps the feature map I, and calculates the vector tracking_id. The vector tracking_id is passed through a linear classifier to obtain an nID-dimensional vector tracking_output, where nID is the number of different targets in the training set. Calculate the nID-dimensional vector target_id based on the true label of the training set; Use cross entropy loss function Loss R e ID Calculate the training loss of the re-ID branch; Loss ReID =CrossEntropyLoss(tracking_id,target_id) CrossEntropyLoss(x,y)={l1,...,l N } T Among them, x is the input, y is the target, ω is the weight, and N is the set minimum batch size.
6. A vehicle detection and tracking device, characterized in that: include: An image extraction module is used to extract the surveillance video to obtain the current frame image, the previous frame image of the current frame, and the heat map of the previous frame image; A feature fusion module, used to extract features from the current frame image, the previous frame image, and the heat map of the previous frame image, and fuse the extracted features to obtain a fused feature map; A target detection module, used to input the fused feature map into a pre-trained target detection and tracking model to detect and track the vehicle; The target detection and tracking model is a model based on CenterTrack, and the backbone network of the target detection and tracking model is DLA-34; the target detection and tracking model includes a first output branch, a second output branch, and a third output branch; the backbone network is used to extract features from the fused feature map, and the feature map extracted by the backbone network is output through the first output branch, the second output branch, and the third output branch to output a target heat map feature map, a target bounding box feature map, and a vehicle target re-identification feature map respectively; After the backbone network extracts features from the fused feature map, the method further includes: processing the extracted feature map based on a spatial attention mechanism to obtain a target feature map; The processing of the fused feature map based on the spatial attention mechanism to obtain a target feature map includes: Performing feature extraction on the fused feature map to obtain a feature map A; Performing maximum pooling and average pooling on the feature map A respectively to obtain feature map A1 and feature map A2, and connecting the feature map A1 and feature map A2 by channel to obtain feature map B; Perform convolution and Sigmoid function processing on the feature map B to obtain a feature map C; Multiply each channel of feature map A and feature map C bit by bit to obtain the target feature map.
