Infrared target detection method, device, equipment and medium
By using pre-trained convolutional neural network model in infrared small object detection, combined with local contrast enhancement unit and channel attention enhancement unit, the problems of poor robustness and limited data-driven methods in the existing technology are solved, and the problem of The precise detection effect of infrared small targets under lower computing resources.
Patent Information
- Application Number
- CN202411597778.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-11
- Publication Date
- 2025-02-18
AI Technical Summary
The existing infrared small object detection methods are poorly robust in complex backgrounds and noise environments, and the data-driven methods are affected by the difficulty of data acquisition and labeling, making them difficult to deploy on edge devices with low computing power.
The pre-trained convolutional neural network model is adopted, combined with the local contrast enhancement unit and the channel attention enhancement unit, and the local contrast attention weight and channel attention weighting processing are used to improve the detection ability of the object detection model for small infrared targets.
Accurate detection of small infrared targets at lower computing resources is achieved, suitable for complex background and noise environments, and is suitable for deployment on resource-constrained edge devices.
Smart Images
Figure CN119478368A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of artificial intelligence technology and relates to an infrared target detection method, device, equipment and medium. Background Art
[0002] Compared with commonly used laser radars, infrared detection equipment does not need to actively emit detection signals, has strong concealment, and can effectively achieve long-distance target positioning, reconnaissance and early warning in typical combat scenarios, thus playing a huge role in complex battlefield environments. Effectively detecting small infrared targets of interest in images has always been an important and challenging task in this field.
[0003] Existing infrared small target detection methods can be roughly divided into two categories: model-driven methods and data-driven methods. Model-driven methods mainly build models manually, do not require a large amount of data support, and usually follow specific assumptions. Model-driven methods are greatly affected by complex backgrounds and noise, and are very sensitive to hyperparameters such as preset window size and segmentation threshold. They need to be repeatedly adjusted and tried according to the characteristics of the scene, and have poor robustness. Data-driven methods mainly rely on deep learning networks, and use supervised learning methods to allow the designed deep learning networks to automatically learn discriminative features from data, and have relatively good generalization in different scenarios. Due to the difficulty in obtaining and high-quality labeling of infrared data, pure data-driven methods are greatly restricted by data, which to a certain extent restricts the further development of the algorithm. At the same time, networks based on CNN and transformer architectures usually have large parameters, making it difficult to deploy them directly on edge computing devices with low computing power. Summary of the invention
[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide an infrared target detection method, device, equipment and medium, which can achieve accurate detection of small infrared targets with relatively low computing resources.
[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions:
[0006] In a first aspect, the present invention provides an infrared target detection method, comprising:
[0007] Inputting the infrared image into an encoder module of a target detection model to obtain a coding feature map, wherein the target detection model is a pre-trained convolutional neural network model;
[0008] Inputting the encoded feature map into a decoder module of a target detection model to obtain a mask of a target to be detected in the infrared image;
[0009] Among them, the encoder module includes a local contrast enhancement unit, which obtains the local contrast attention weight of the pixel points in the infrared image based on the local contrast distance of the pixel points in the infrared image, and fuses the local contrast attention weight with the feature map of the infrared image.
[0010] Further, based on the local contrast distance of the pixel points in the infrared image, obtaining the local contrast attention weight of the pixel points in the infrared image includes:
[0011] Using 4 expansion rates , size is The convolution operator As a window, calculate the local contrast distance of each pixel in the infrared image and obtain 4 local feature maps ;
[0012] Among them, the pixel point in the infrared image The calculation formula of the local contrast distance is:
[0013] ,
[0014] ,
[0015] ,
[0016] ,
[0017] in, , , and Pixel 4 local contrast distances; Represents the pixel point in the infrared image Gray value of and is a hyperparameter greater than 0;
[0018] The pixel point The local contrast attention weight is:
[0019] ,
[0020] in, express Activation function.
[0021] Furthermore, the local contrast enhancement unit further includes: Convolutional layers, depthwise separable convolutional layers, and batch normalization layers;
[0022] Through the The convolution layer expands the channels of the infrared image to obtain the expanded Channel feature map;
[0023] The local contrast attention weights are fused with the expanded feature map through the depthwise separable convolutional layer and batch normalization layer, including:
[0024] Based on the local feature map , through the sigmoid activation function, the local contrast attention weight matrix of the infrared image is obtained; the local contrast attention weight matrix The calculation formula is:
[0025] ,
[0026] in, represents element-by-element multiplication, It means adding elements one by one;
[0027] The local contrast attention weight matrix With the expanded The channel feature map is dot-multiplied element by element to obtain the first feature map;
[0028] Inputting the first feature map into the depthwise separable convolution layer and the batch normalization layer, performing depthwise separable convolution processing and batch normalization processing, to obtain a second feature map;
[0029] The expanded The channel feature map and the second feature map are added element by element and then activated by the PReLU function to obtain the third feature map.
[0030] Furthermore, the encoder module also includes residual layers, the The residual layers are connected in sequence, and the branches of each residual layer are directly output;
[0031] Through the A residual layer is used to extract features from the third feature map to obtain Encoded feature maps of different depths.
[0032] Furthermore, the target detection model also includes a reinforcement module, which includes Channel attention reinforcement unit;
[0033] Through the channel attention strengthening units, respectively The encoded feature maps are weighted by attention and we get An enhanced feature map.
[0034] Furthermore, the channel attention enhancement unit includes a global average pooling layer, a one-dimensional convolution layer and a Sigmod activation function layer;
[0035] Through the channel attention enhancement unit, the corresponding encoding feature map is subjected to attention weighting processing to obtain an enhanced feature map, including:
[0036] The encoded feature map is sequentially input into the global average pooling layer, the one-dimensional convolution layer and the Sigmod activation function layer to obtain the weight value of each channel of the encoded feature map;
[0037] Multiplying the weight value of each channel of the encoding feature map by the encoding feature map point by point to obtain a weighted feature map;
[0038] The weighted feature map is added to the encoded feature map to obtain an enhanced feature map.
[0039] Furthermore, the decoder module comprises Upsampling fusion layer;
[0040] Through the Upsampling fusion layer, the different depths The enhanced feature maps are upsampled to align the enhanced feature maps, and the aligned enhanced feature maps are subjected to feature fusion.
[0041] In a second aspect, the present invention further provides an infrared target detection device, the device comprising:
[0042] An encoding processing module, used for inputting the infrared image into an encoder module of a target detection model to obtain an encoding feature map, wherein the target detection model is a pre-trained convolutional neural network model;
[0043] A decoding processing module, used for inputting the encoded feature map into a decoder module of a target detection model to obtain a mask of a target to be detected in an infrared image;
[0044] Among them, the encoder module includes a local contrast enhancement unit, which obtains the local contrast attention weight of the pixel points in the infrared image based on the local contrast distance of the pixel points in the infrared image, and fuses the local contrast attention weight with the feature map of the infrared image.
[0045] In a third aspect, the present invention further provides a computer device, comprising:
[0046] Memory for storing computer programs;
[0047] A processor is used to execute the computer program to implement the steps of the above-mentioned infrared target detection method.
[0048] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the steps of the above-mentioned infrared target detection method.
[0049] Compared with the prior art, the present invention has the following beneficial effects:
[0050] The infrared target detection method provided by the present invention calculates the local contrast distance of the pixel points in the infrared image through four convolution operators, obtains the local contrast attention weight of the pixel points in the infrared image, guides the neural network to strengthen the attention to the spatial position of the potential small target during the encoding process, so that the target detection model is adapted to the detection of infrared small targets; and realizes the efficient fusion of features between different channels through the channel attention strengthening unit, strengthens the attention to the potential small targets in different channels, and is suitable for the detection of infrared small targets. The target detection model designed by the present invention has relatively few parameters and calculations, and can realize accurate detection of infrared small targets with low computing resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 A schematic diagram of a flow chart of an infrared target detection method provided by an embodiment of the present invention;
[0052] Figure 2 A schematic diagram of the network structure of a target detection model in an embodiment of the present invention;
[0053] Figure 3 is a schematic diagram of a network structure of a local contrast enhancement unit in an embodiment of the present invention;
[0054] Figure 4 is a schematic diagram of a network structure of a local contrast enhancement layer in an embodiment of the present invention;
[0055] Figure 5 is a schematic diagram of connecting multiple residual layers in an embodiment of the present invention;
[0056] Figure 6 A schematic diagram of the network structure of a channel attention enhancement unit according to an embodiment of the present invention;
[0057] Figure 7 A schematic diagram of the network structure of an upsampling fusion layer according to an embodiment of the present invention;
[0058] Figure 8 A schematic diagram of the network structure of a prediction head according to an embodiment of the present invention;
[0059] Fig. 9 A schematic diagram of the structure of an infrared target detection device provided by an embodiment of the present invention;
[0060] Fig.10 An internal structure diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0061] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. The same reference numerals in the accompanying drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. The embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.
[0062] The term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: 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.
[0063] Embodiment 1:
[0064] like Figures 1 to 8 As shown, an embodiment of the present invention provides an infrared target detection method. Figure 1 The flowchart is a flow chart of the infrared target detection method. This flowchart only shows the logical sequence of the method described in this embodiment. Under the premise of no conflict, in other possible embodiments of the present invention, different Figure 1 The steps shown or described are accomplished in the order shown.
[0065] The infrared target detection method provided in this embodiment can be applied to a terminal, and can be executed by an infrared target detection device. The device can be implemented by software and / or hardware, and the device can be integrated in the terminal.
[0066] See also Figure 1 The method of the embodiment of the present invention specifically includes the following steps:
[0067] Step 1: Input the infrared image into the encoder module of the target detection model to obtain a coding feature map, where the target detection model is a pre-trained convolutional neural network model;
[0068] Among them, the training images and test images used for training the target detection model in the embodiment of the present invention use three public infrared datasets: NUAA-SIRST, NUDT-SIRST, and IRSTD-1K. Before training, the data of the three datasets are preprocessed, and the specific steps are: standardizing the images, adjusting the size to 256×256 pixels by edge filling and random cropping, and using random flipping for data enhancement; then dividing the preprocessed infrared image dataset into a training set and a test set according to the built-in method of the dataset.
[0069] The present invention uses the soft intersection over union (softIoU) loss function for supervised training to adjust the parameters of the target detection model. The number of training iterations is 400, the initial learning rate is set to 0.0005, and the weight decay is 10 -8 , a polynomial decay learning rate scheduling strategy was adopted, the Adam optimizer was used, the learning rate decay rate was set to 0.1, and the number of adjustment iterations was 200 and 300 respectively.
[0070] like Figure 2 As shown, the target detection model of the present invention includes an encoder module, an enhancement module and a decoder module.
[0071] Wherein, the encoder module includes a local contrast enhancement unit, Figure 3 The schematic diagram of the network structure of the local contrast enhancement unit is shown, including the local contrast enhancement layer, Convolutional layers, depthwise separable convolutional layers, and batch normalization layers.
[0072] Figure 4 The network structure diagram of the local contrast enhancement layer is shown in FIG. , size is The convolution operator As a window, the local contrast distance of each pixel in the infrared image is calculated, and the local contrast distance of each pixel is combined to obtain 4 local feature maps .
[0073] In the embodiment of the present invention, the expansion rate Take 1, size Take 3.
[0074] Based on the local feature map , through the sigmoid activation function, the local contrast attention weight matrix of the infrared image is obtained. The local contrast attention weight matrix The calculation formula is:
[0075] ,
[0076] in, represents element-by-element multiplication, It represents element-by-element addition.
[0077] Specifically, the local contrast distance and local contrast attention weight of each pixel in the infrared image are calculated by For example, the calculation formula of its local contrast distance is:
[0078] ,
[0079] ,
[0080] ,
[0081] ,
[0082] in, , , and Pixel 4 local contrast distances; Represents the pixel point in the infrared image Gray value of Take 1, Take 0.5.
[0083] The pixel point The calculation formula of the local contrast attention weight is:
[0084] .
[0085] Through the The convolution layer expands the channels of the infrared image to obtain the expanded Channel feature map ,in , and are the number of channels, height and width of the expanded feature map respectively.
[0086] Through the depth-separable convolution layer and batch normalization layer, the local contrast attention weight matrix With the expanded Channel feature map Perform element-by-element dot multiplication to obtain the first feature map ;
[0087] The first feature map Input the depth separable convolution layer and batch normalization layer, perform depth separable convolution processing and batch normalization processing, and obtain the second feature map ;
[0088] The expanded Channel feature map And the second feature map After adding each element one by one, the third feature map is obtained by PReLU activation function. .
[0089] like Figure 5 As shown, the encoder module of the present invention also includes residual layers, the The residual layers are connected in sequence, and the branches of each residual layer are directly output. residual layer, for the third feature map Perform feature extraction to obtain Encoded feature maps of different depths.
[0090] In an embodiment of the present invention, , for the third feature map of the input , after each residual layer is processed, an encoded feature map is obtained ,in , and They are respectively the encoding feature maps The number of channels, height, and width.
[0091] The target detection model further includes an enhancement module, which includes channel attention enhancement units equal to the number of residual layers. Therefore, in the embodiment of the present invention, the enhancement module has four channel attention enhancement units. The four encoding feature maps output by the encoder module are respectively Perform attention weighted processing to obtain 4 weighted feature maps , and the input original encoding feature map With weighted results Fusion is performed at the spatial level, and the enhanced feature map is finally output .
[0092] The channel attention enhancement unit can realize the efficient fusion of information between different channels and enhance the attention on potential small infrared target areas.
[0093] like Figure 6 As shown, the channel attention enhancement unit includes a global average pooling layer, a one-dimensional convolution layer, and a Sigmod activation function layer. An encoded feature map is input into the global average pooling layer, the one-dimensional convolution layer, and the Sigmod activation function layer to obtain the weight value of each channel thereof; the weight value of each channel of the encoded feature map is multiplied point by point with the encoded feature map to obtain a weighted feature map; the weighted feature map is added to the encoded feature map to obtain an enhanced feature map.
[0094] Step 2: Input the encoded feature map into the decoder module of the target detection model to obtain the mask of the target to be detected in the infrared image.
[0095] The decoder module of the present invention consists of It consists of an upsampling fusion layer and a prediction head.
[0096] Through the Upsampling fusion layer, the different depths The enhanced feature maps are upsampled to align the enhanced feature maps, and the aligned enhanced feature maps are subjected to feature fusion.
[0097] Figure 7 The network structure of the upsampling fusion layer is shown. The upsampling fusion layer strengthens the input feature map Upsample the low-level feature maps in the layer, increase the spatial resolution while reducing the number of channels, and make it consistent with the feature maps of the current layer After alignment, the two are added together to achieve feature fusion and reconstruction.
[0098] Figure 8 The network structure of the prediction head is shown. Through the prediction head, the target prediction results in the infrared image are output. , which is the mask of the target to be detected in the infrared image.
[0099] The infrared image data to be inferred is sent to the trained target detection model to obtain the pixel-level detection mask result of small targets in the infrared image.
[0100] Table 1 shows the comparison results of the target detection model of the present invention with other advanced algorithm models in terms of normalized intersection over union (nIoU), detection probability (Pd) and false alarm probability (Fa). Table 2 lists the comparison of other methods in Table 1 that are close to the target detection model results of the present invention in terms of model parameter quantity and floating point operation quantity. It can be seen that compared with other advanced algorithm models, the target detection model proposed by the present invention has excellent performance and fewer parameters, and can well balance the requirements of computational complexity and detection performance, and is more suitable for implementation on resource-constrained edge infrared detection equipment.
[0101] Table 1
[0102]
[0103] Table 2
[0104]
[0105] Embodiment 2:
[0106] Based on the same inventive concept as in Example 1, the embodiment of the present invention further provides an infrared target detection device for implementing the above infrared target detection method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the infrared target detection device embodiment provided below can refer to the limitations of the infrared target detection method above, and will not be repeated here.
[0107] like Fig. 9As shown, an embodiment of the present invention provides an infrared target detection device, comprising:
[0108] An encoding processing module, used for inputting the infrared image into an encoder module of a target detection model to obtain an encoding feature map, wherein the target detection model is a pre-trained convolutional neural network model;
[0109] A decoding processing module, used for inputting the encoded feature map into a decoder module of a target detection model to obtain a mask of a target to be detected in an infrared image;
[0110] Among them, the encoder module includes a local contrast enhancement unit, which obtains the local contrast attention weight of the pixel points in the infrared image based on the local contrast distance of the pixel points in the infrared image, and fuses the local contrast attention weight with the feature map of the infrared image.
[0111] Embodiment 3:
[0112] The embodiment of the present invention further provides a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Fig.10 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps of the infrared target detection method in the aforementioned embodiment are implemented.
[0113] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0114] Embodiment 4:
[0115] The embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the following method are implemented:
[0116] Inputting the infrared image into an encoder module of a target detection model to obtain a coding feature map, wherein the target detection model is a pre-trained convolutional neural network model;
[0117] Inputting the encoded feature map into a decoder module of a target detection model to obtain a mask of a target to be detected in the infrared image;
[0118] Among them, the encoder module includes a local contrast enhancement unit, which obtains the local contrast attention weight of the pixel points in the infrared image based on the local contrast distance of the pixel points in the infrared image, and fuses the local contrast attention weight with the feature map of the infrared image.
[0119] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products, and therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0120] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0121] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0123] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.
Claims
1. An infrared target detection method, characterized in that: include: Inputting the infrared image into an encoder module of a target detection model to obtain a coding feature map, wherein the target detection model is a pre-trained convolutional neural network model; Inputting the encoded feature map into a decoder module of a target detection model to obtain a mask of a target to be detected in the infrared image; Among them, the encoder module includes a local contrast enhancement unit, which obtains the local contrast attention weight of the pixel points in the infrared image based on the local contrast distance of the pixel points in the infrared image, and fuses the local contrast attention weight with the feature map of the infrared image.
2. The infrared target detection method according to claim 1, characterized in that: Based on the local contrast distance of the pixel points in the infrared image, obtaining the local contrast attention weight of the pixel points in the infrared image includes: Using 4 expansion rates , size is The convolution operator As a window, calculate the local contrast distance of each pixel in the infrared image and obtain 4 local feature maps ; Among them, the pixel points in the infrared image The calculation formula of the local contrast distance is: ， ， ， ， in, , , and Pixel 4 local contrast distances; Represents the pixel point in the infrared image Gray value of and is a hyperparameter greater than 0; The pixel point The local contrast attention weight is: ， in, express Activation function.
3. The infrared target detection method according to claim 2, characterized in that: The local contrast enhancement unit also includes: Convolutional layers, depthwise separable convolutional layers, and batch normalization layers; Through the The convolution layer expands the channels of the infrared image to obtain the expanded Channel feature map; The local contrast attention weights are fused with the expanded feature map through the depthwise separable convolutional layer and batch normalization layer, including: Based on the local feature map , through the sigmoid activation function, the local contrast attention weight matrix of the infrared image is obtained; the local contrast attention weight matrix The calculation formula is: ， in, represents element-by-element multiplication, It means adding elements one by one; The local contrast attention weight matrix With the expanded The channel feature map is dot-multiplied element by element to obtain the first feature map; Inputting the first feature map into the depthwise separable convolution layer and the batch normalization layer, performing depthwise separable convolution processing and batch normalization processing, to obtain a second feature map; The expanded The channel feature map and the second feature map are added element by element and then activated by the PReLU function to obtain the third feature map.
4. The infrared target detection method according to claim 3, characterized in that: The encoder module also includes residual layers, the The residual layers are connected in sequence, and the branches of each residual layer are directly output; Through the A residual layer is used to extract features from the third feature map to obtain Encoded feature maps of different depths.
5. The infrared target detection method according to claim 4, characterized in that: The target detection model also includes a reinforcement module, which includes Channel attention reinforcement unit; Through the channel attention strengthening units, respectively The encoded feature maps are weighted by attention and we get An enhanced feature map.
6. The infrared target detection method according to claim 5, characterized in that: The channel attention enhancement unit includes a global average pooling layer, a one-dimensional convolution layer and a Sigmod activation function layer; Through the channel attention enhancement unit, the corresponding encoding feature map is subjected to attention weighting processing to obtain an enhanced feature map, including: The encoded feature map is sequentially input into the global average pooling layer, the one-dimensional convolution layer and the Sigmod activation function layer to obtain the weight value of each channel of the encoded feature map; Multiplying the weight value of each channel of the encoded feature map by the encoded feature map point by point to obtain a weighted feature map; The weighted feature map is added to the encoded feature map to obtain an enhanced feature map.
7. The infrared target detection method according to claim 5 or 6, characterized in that: The decoder module includes Upsampling fusion layer; Through the Upsampling fusion layer, the different depths The enhanced feature maps are upsampled to align the enhanced feature maps, and the aligned enhanced feature maps are subjected to feature fusion.
8. An infrared target detection device, characterized in that: include: An encoding processing module, used for inputting the infrared image into an encoder module of a target detection model to obtain an encoding feature map, wherein the target detection model is a pre-trained convolutional neural network model; A decoding processing module, used for inputting the encoded feature map into a decoder module of a target detection model to obtain a mask of a target to be detected in an infrared image; Among them, the encoder module includes a local contrast enhancement unit, which obtains the local contrast attention weight of the pixel points in the infrared image based on the local contrast distance of the pixel points in the infrared image, and fuses the local contrast attention weight with the feature map of the infrared image.
9. A computer device, characterized in that: include: Memory for storing computer programs; A processor, configured to execute the computer program to implement the steps of the infrared target detection method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the infrared target detection method described in any one of claims 1 to 7 are implemented.