A Remote Sensing Image Aircraft Positioning Method Based on a Full-Channel Spatial Attention Module

Abstract

The present invention discloses a method for aircraft positioning in remote sensing images based on a full-channel spatial attention module, comprising the following steps: S1: Obtain the remote sensing image to be positioned; S2: After preprocessing the remote sensing image to be positioned, input it into a pre-trained remote sensing image aircraft positioning model to obtain the aircraft positioning result in the remote sensing image to be positioned; wherein, the remote sensing image aircraft positioning model includes a convolutional neural network, a full-channel spatial attention module, and a target detection and positioning module connected in sequence. The present invention has higher detection accuracy, better positioning accuracy, lower computational resource requirements, and higher applicability.

Description

Technical Field

[0001] The present invention relates to the technical field of target detection and positioning, and more particularly to a method for positioning airplanes in remote sensing images based on a full-channel spatial attention module. Background Art

[0002] Remote sensing image analysis has important application values in fields such as geographic information system (GIS), environmental monitoring, disaster assessment, agricultural management, and national defense. In this field, small target detection and positioning technologies are important tasks in remote sensing image processing, especially when accurately identifying and positioning small targets such as small airplanes, which has extensive practical significance.

[0003] However, due to the high complexity of remote sensing images, existing small target detection and positioning methods still have problems such as low detection accuracy, poor positioning accuracy, high computational resource requirements, and low applicability.

[0004] Therefore, how to provide a method for positioning airplanes in remote sensing images with higher detection accuracy, better positioning accuracy, lower computational resource requirements, and higher applicability is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method for positioning airplanes in remote sensing images based on a full-channel spatial attention module.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A method for positioning airplanes in remote sensing images based on a full-channel spatial attention module includes the following steps:

[0008] S1: Obtain the remote sensing image to be positioned;

[0009] S2: After preprocessing the remote sensing image to be positioned, input it into a pre-trained remote sensing image airplane positioning model to obtain the airplane positioning result in the remote sensing image to be positioned; wherein, the remote sensing image airplane positioning model includes a convolutional neural network, a full-channel spatial attention module, and a target detection and positioning module connected in sequence.

[0010] Preferably, the convolutional neural network includes a convolutional module / includes a plurality of convolutional modules connected in sequence; wherein, the convolutional module includes a convolutional layer, a non-linear activation function layer, a pooling layer, and a batch normalization layer connected in sequence.

[0011] Preferably, the all-channel spatial attention module includes an adaptive average pooling layer, a first 1×1 convolutional layer, a first activation function layer, a second 1×1 convolutional layer, an adaptive max pooling layer, a third 1×1 convolutional layer, a second activation function layer, a fourth 1×1 convolutional layer, an element-wise addition unit, a third activation function layer, a first element-wise multiplication unit, a channel maximum value processing unit, a channel average value processing unit, a channel concatenation layer, a 7×7 convolutional layer, a fourth activation function layer, and a second element-wise multiplication unit;

[0012] The input ends of the adaptive average pooling layer and the adaptive max pooling layer are both connected to the output end of the convolutional neural network;

[0013] The output end of the adaptive average pooling layer is sequentially connected to the input end of the element-wise addition unit through the first 1×1 convolutional layer, the first activation function layer, and the second 1×1 convolutional layer;

[0014] The output end of the adaptive max pooling layer is sequentially connected to the input end of the element-wise addition unit through the third 1×1 convolutional layer, the second activation function layer, and the fourth 1×1 convolutional layer;

[0015] The output end of the element-wise addition unit is connected to the input end of the first element-wise multiplication unit through the third activation function layer;

[0016] The output end of the convolutional neural network is connected to the input end of the first element-wise multiplication unit;

[0017] The output end of the first element-wise multiplication unit is respectively connected to the input ends of the channel maximum value processing unit and the channel average value processing unit;

[0018] The output ends of the channel maximum value processing unit and the channel average value processing unit are both connected to the input end of the channel concatenation layer;

[0019] The output end of the channel concatenation layer is sequentially connected to the input end of the second element-wise multiplication unit through the 7×7 convolutional layer and the fourth activation function layer;

[0020] The output end of the first element-wise multiplication unit is connected to the input end of the second element-wise multiplication unit;

[0021] The output end of the second element-wise multiplication unit is connected to the input end of the target detection and localization module.

[0022] Preferably, the target detection and localization module includes a convolutional layer, a non-linear activation layer, a downsampling layer, and an upsampling layer that are sequentially connected;

[0023] The input end of the convolutional layer of the target detection and positioning module is connected to the output end of the second element-wise multiplication unit;

[0024] The output end of the upsampling layer is the output end of the remote sensing image aircraft positioning model.

[0025] Preferably, the preprocessing in S2 includes scaling processing, normalization processing, and denoising processing.

[0026] Preferably, the pre-trained remote sensing image aircraft positioning model is obtained based on the following method:

[0027] S21: Obtain an initial remote sensing image dataset; wherein, the initial remote sensing image dataset includes a number of remote sensing images containing aircraft;

[0028] S22: Perform data annotation on the initial remote sensing image dataset; wherein, the annotation content includes the bounding box of the aircraft, the coordinates of the bounding box, and the category of the bounding box;

[0029] S23: Perform preprocessing on the annotated initial remote sensing image dataset to obtain a final remote sensing image dataset;

[0030] S24: Input the final remote sensing image dataset into the remote sensing image aircraft positioning model for training to obtain a pre-trained remote sensing image aircraft positioning model.

[0031] Preferably, the preprocessing in S23 includes scaling processing, normalization processing, data augmentation processing, and denoising processing.

[0032] Preferably, the remote sensing image aircraft positioning model is trained based on the following total loss function:

[0033] L total = μL t + vL m + ηL s ;

[0034] wherein, L total represents the total loss; L t , L m , L s represent the object loss, the feature enhancement consistency loss, and the attention sparsity loss in sequence; μ, v, η represent weight coefficients.

[0035] Preferably, L t = αL box + βL cls ;

[0036]

[0037] L s = γ1L channel+γ2L spatial ;

[0038] wherein, L box represents the bounding box regression loss; L cls represents the classification loss; α and β represent the weight coefficients; X O represents the original features output by the convolutional neural network; X O [i] represents the i-th pixel of X O ; X en represents the enhanced features output by the full-channel spatial attention module; X en [i] represents the i-th pixel of X en ; N represents the number of pixels included in X O and X en ; L channel represents the channel attention sparsity loss; L spatial represents the spatial attention sparsity loss; γ1 and γ2 represent hyperparameters; ||X O [i] - X en [i]|| 2 represents the square of the Euclidean norm of X O [i] - X en [i].

[0039] Preferably, L box = 1 - GIoU(B true , B pred );

[0040] L cls = -y c log(p c );

[0041]

[0042]

[0043] wherein, GIoU represents the overlap degree between the predicted bounding box and the ground truth bounding box; B true represents the annotated ground truth bounding box; B pred represents the predicted bounding box; y c represents the one-hot encoding of the class of the ground truth bounding box; p c represents the classification probability of the predicted bounding box; M c represents the channel attention map output by the third activation function layer; ||M c ||1 represents the L1-norm of M c ; M c [i, j] represents the pixel at the i-th row and j-th column of M c ; M s represents the spatial attention map output by the fourth activation function layer; ||Ms || 1 represents the L1-norm of M s ; M s [i, j] represents the pixel at the i-th row and j-th column of M s ; H represents the height of M c and M s ; W represents the width of M c and M s .

[0044] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method for aircraft positioning in remote sensing images based on a full-channel spatial attention module, which can obtain the following beneficial technical effects:

[0045] 1) Improved detection accuracy and positioning accuracy: By introducing the full-channel spatial attention module, the present invention significantly enhances the ability to capture small target features, making the model perform more excellently in detecting and positioning small aircraft in remote sensing images. At the same time, the design of the total loss function effectively balances the target positioning and classification tasks, further improving the detection accuracy and robustness. Compared with traditional methods, the present invention is particularly prominent in dealing with small target detection in complex backgrounds, lighting changes or occlusion situations, and the positioning error is significantly reduced.

[0046] 2) Low computational resource requirements and strong applicability: The full-channel spatial attention module adopts a lightweight design and can complete the adaptive enhancement of features through a small amount of parameter calculations, effectively reducing the computational overhead. Combined with data preprocessing and noise removal techniques, the present invention maintains low resource requirements while improving the model performance, enabling the method of the present invention to be deployed in various computing environments. In addition, the full-channel spatial attention module provided by the present invention has strong versatility and can be flexibly embedded in existing deep learning network architectures, applicable to small target detection tasks of different scales, and can be widely applied to remote sensing image analysis, aviation detection and other related fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0048] Figure 1 is a schematic structural diagram of the aircraft positioning model for remote sensing images provided in the embodiment of the present invention;

[0049] Figure 2 is a schematic structural diagram of the full-channel spatial attention module provided in the embodiment of the present invention;

[0050] Figure 3 This is a schematic diagram of the structure of the target detection and positioning module provided in the embodiments of the present invention. Detailed implementation manners

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0052] The embodiments of the present invention disclose a method for positioning an aircraft in a remote sensing image based on a full-channel spatial attention module, including the following steps:

[0053] S1: Obtain the remote sensing image to be positioned;

[0054] S2: After preprocessing the remote sensing image to be positioned, input it into a pre-trained remote sensing image aircraft positioning model to obtain the aircraft positioning result in the remote sensing image to be positioned; wherein, the remote sensing image aircraft positioning model includes a convolutional neural network, a full-channel spatial attention module, and a target detection and positioning module connected in sequence (as Figure 1 shown).

[0055] It should be noted that: the aircraft positioning result in the remote sensing image to be positioned includes the predicted bounding box of the aircraft, the coordinates of the predicted bounding box, the category of the predicted bounding box, and the classification probability of the predicted bounding box.

[0056] In one embodiment, the convolutional neural network includes a convolutional module / includes a plurality of convolutional modules connected in sequence; wherein, the convolutional module includes a convolutional layer, a non-linear activation function layer, a pooling layer, and a batch normalization layer connected in sequence.

[0057] In one embodiment, as Figure 2 shown, the full-channel spatial attention module includes an adaptive average pooling layer, a first 1*1 convolutional layer, a first activation function layer, a second 1*1 convolutional layer, an adaptive max pooling layer, a third 1*1 convolutional layer, a second activation function layer, a fourth 1*1 convolutional layer, an element-wise addition unit, a third activation function layer, a first element-wise multiplication unit, a channel maximum value processing unit, a channel average value processing unit, a channel concatenation layer, a 7*7 convolutional layer, a fourth activation function layer, and a second element-wise multiplication unit;

[0058] The input ends of the adaptive average pooling layer and the adaptive max pooling layer are both connected to the output end of the convolutional neural network;

[0059] The output end of the adaptive average pooling layer is successively connected to the input end of the first 1×1 convolutional layer, the first activation function layer, the second 1×1 convolutional layer and the element-wise addition unit;

[0060] The output end of the adaptive max pooling layer is successively connected to the input end of the third 1×1 convolutional layer, the second activation function layer, the fourth 1×1 convolutional layer and the element-wise addition unit;

[0061] The output end of the element-wise addition unit is connected to the input end of the third activation function layer and then to the input end of the first element-wise multiplication unit;

[0062] The output end of the convolutional neural network is connected to the input end of the first element-wise multiplication unit;

[0063] The output end of the first element-wise multiplication unit is respectively connected to the input ends of the channel maximum value processing unit and the channel average value processing unit;

[0064] The output ends of the channel maximum value processing unit and the channel average value processing unit are both connected to the input end of the channel splicing layer;

[0065] The output end of the channel splicing layer is successively connected to the input end of the 7×7 convolutional layer, the fourth activation function layer and the second element-wise multiplication unit;

[0066] The output end of the first element-wise multiplication unit is connected to the input end of the second element-wise multiplication unit;

[0067] The output end of the second element-wise multiplication unit is connected to the input end of the target detection and localization module.

[0068] It can be understood that:

[0069] Each pixel in the feature map output by the channel maximum value processing unit takes the maximum value among the pixels at the corresponding positions in the channel feature maps output by the first element-wise multiplication unit;

[0070] Each pixel in the feature map output by the channel average value processing unit takes the average value among the pixels at the corresponding positions in the channel feature maps output by the first element-wise multiplication unit;

[0071] The first 1×1 convolutional layer and the third 1×1 convolutional layer share weights;

[0072] The first activation function layer and the second activation layer share weights;

[0073] The second 1×1 convolutional layer and the fourth 1×1 convolutional layer share weights;

[0074] Both the first activation function layer and the second activation function layer are Relu functions;

[0075] Both the third activation function layer and the fourth activation function layer are Sigmoid functions.

[0076] In one embodiment, as Figure 3 shown, the target detection and positioning module includes a convolutional layer, a non-linear activation layer, a downsampling layer, and an upsampling layer connected in sequence;

[0077] The input end of the convolutional layer of the target detection and positioning module is connected to the output end of the second element-wise multiplication unit;

[0078] The output end of the upsampling layer is the output end of the remote sensing image aircraft positioning model.

[0079] In one embodiment, the preprocessing in S2 includes scaling processing, normalization processing, and denoising processing (the same as the scaling processing, normalization processing, and denoising processing used in the training stage).

[0080] In one embodiment, a pre-trained remote sensing image aircraft positioning model is obtained based on the following method:

[0081] S21: Obtain an initial remote sensing image dataset; wherein, the initial remote sensing image dataset includes a number of remote sensing images containing aircraft;

[0082] It can be understood that: in order to ensure that the remote sensing image aircraft positioning model can handle various different scenarios, thereby improving its generalization ability. The present invention collects remote sensing images from multiple sources, including but not limited to satellite images, aerial photography, and images taken by drones. These images cover different geographical regions, such as cities, villages, mountains, and oceans, etc., to ensure that the model can work in a diverse environment. At the same time, the collected images also include different weather conditions, such as clear, cloudy, rainy, and snowy days, as well as different time periods, such as morning, noon, and evening, to simulate various situations where aircraft may appear in the real world.

[0083] S22: Perform data annotation on the initial remote sensing image dataset; wherein, the annotation content includes the bounding box of the aircraft, the coordinates of the bounding box, and the category of the bounding box;

[0084] It can be understood that: the present invention uses professional image annotation tools to accurately annotate the collected remote sensing images. The main task of annotation is to identify the aircraft in the image and draw a bounding box for it. The drawing of the bounding box needs to follow certain rules, using a rectangular box to closely fit the shape of the aircraft to ensure that every part of the aircraft is included. In addition, the annotation content also includes the coordinates of the bounding box and the category of the bounding box (the category specifically refers to whether it is an aircraft).

[0085] S23: Preprocess the labeled initial remote sensing image dataset to obtain the final remote sensing image dataset;

[0086] In one embodiment, the preprocessing in S23 includes scaling processing, normalization processing, data augmentation processing, and denoising processing.

[0087] Specifically:

[0088] 1) The scaling processing can be expressed as: O' = Resize(O, H', W ′ )

[0089] where O represents the remote sensing image before scaling, and O' represents the remote sensing image after scaling. H', W' represent the height and width of the remote sensing image after scaling;

[0090] 2) The normalization processing can be expressed as:

[0091] where 0″ is the remote sensing image after normalization; the normalization processing means scaling the pixel values from [0, 255] to the interval [0, 1].

[0092] 3) The data augmentation processing includes operations such as random rotation, horizontal flipping, vertical flipping, and random cropping;

[0093] The random rotation, horizontal flipping, and vertical flipping operations can be expressed as: O R = Rotate(O", θ) where θ represents the randomly selected rotation angle, and O R represents the remote sensing image after rotation.

[0094] 4) The denoising processing can be expressed as: O denoise = Denoise(O R , σ)

[0095] where σ represents the standard deviation of the filter, and O denoise represents the remote sensing image after denoising. The noise removal algorithm specifically applied in the present invention can be Gaussian filtering, median filtering, etc.

[0096] Through the above preprocessing steps, a series of high-quality, diverse, and model-training-suitable image data can be obtained. These preprocessing operations not only improve the efficiency of model training but also enhance the accuracy and robustness of the model for aircraft positioning under different conditions.

[0097] S24: Input the final remote sensing image dataset into the remote sensing image aircraft positioning model for training to obtain a pre-trained remote sensing image aircraft positioning model.

[0098] Specifically:

[0099] 1. Input the denoised image into a convolutional neural network to obtain the original feature X O ;

[0100] 2. Input the original feature X O into the global average pooling layer to obtain the first-channel description vector X avg ;

[0101] Input the original feature X O into the global max pooling layer to obtain the second-channel description vector X max ;

[0102] 3. Input the first-channel description vector X avg sequentially into the first 1×1 convolutional layer, the first activation function layer, and the second 1×1 convolutional layer to obtain the intermediate feature map X1;

[0103] 4. Input the second-channel description vector X max sequentially into the third 1×1 convolutional layer, the second activation function layer, and the fourth 1×1 convolutional layer to obtain the intermediate feature map X2;

[0104] 5. Add the intermediate feature map X1 and the intermediate feature map X2 element-wise, and then input the result into the third activation function layer to obtain the channel attention map M c ;

[0105] 6. Multiply the channel attention map M c element-wise with the original feature X O to obtain the channel-enhanced feature X ch ;

[0106] 7. Take the maximum value of each channel in the channel-enhanced feature X ch to obtain the intermediate feature map X3;

[0107] Take the average value of each channel in the channel-enhanced feature X ch to obtain the intermediate feature map X4;

[0108] It can be understood that:

[0109] Each pixel in the intermediate feature map X3 takes the maximum value of the corresponding pixels in each channel of X ch ;

[0110] Each pixel in the intermediate feature map X4 takes the average value of the corresponding pixels in each channel of X ch ;

[0111] 8. Concatenate the intermediate feature map X3 and the intermediate feature map X4 in the channel dimension and then input them into the 7×7 convolutional layer and the fourth activation function layer in sequence to obtain the spatial attention map M s ;

[0112] 9. Multiply the spatial attention map M s and the feature X after channel enhancement ch element-wise to obtain the enhanced feature X en .

[0113] 10. Input the enhanced feature X en into the convolutional layer, non-linear activation layer, downsampling layer and upsampling layer in the target detection and localization module in sequence to obtain the aircraft localization result, where the aircraft localization result includes the predicted bounding box, the coordinates of the predicted bounding box, the category of the predicted bounding box, and the classification probability of the predicted bounding box (i.e., the confidence that the target in the predicted bounding box is an aircraft).

[0114] It can be understood that: the present invention uses the non-maximum suppression technology to screen out the bounding boxes with higher confidence and non-overlapping from the predicted bounding boxes output by the model, remove the redundant detection results, and obtain the aircraft localization result.

[0115] It can be understood that: the present invention can also map the position of the predicted bounding box back to the coordinate system of the original remote sensing image according to the coordinates (x min , y min , x max , y max ) of the predicted bounding box (where (x min , y min ) represents the coordinates of the upper left corner of the predicted bounding box; (x max , y max ) represents the coordinates of the lower right corner of the predicted bounding box). Then, draw the predicted bounding box, the category of the predicted bounding box, and the classification probability of the predicted bounding box (i.e., the confidence that the target in the predicted bounding box is an aircraft) on the original remote sensing image.

[0116] In one embodiment, the remote sensing image aircraft localization model is trained based on the following total loss function:

[0117] L total = μL t + vL m + ηL s ;

[0118] where, L total represents the total loss; L t , L m , L sThey represent target loss, feature enhancement consistency loss, and attention sparsity loss respectively; μ, v, η represent weight coefficients.

[0119] In one embodiment, L t =αL box +βL cls ;

[0120]

[0121] L s =γ1L channel +γ2L spatial ;

[0122] Among them, L box represents the bounding box regression loss; L cls represents classification loss; α and β represent weight coefficients; X O Represents the original features of the convolutional neural network output; X O [i] indicates X O The i-th pixel of en represents the enhanced features output by the full-channel spatial attention module; X en [i] indicates X en The i-th pixel of O and X en The number of pixels included; L channel represents the channel attention sparsity loss; L spatial represents the spatial attention sparsity loss; γ1 and γ2 represent hyperparameters used to balance the channel attention sparsity loss and the spatial attention sparsity loss; ||X O [i]-X en [i]|| 2 Represents X O [i]-X en The square of the Euclidean norm of [i].

[0123] In one embodiment, L box =1-GIoU(B true , B pred );

[0124] L cls =-y c log(p c );

[0125]

[0126]

[0127] Among them, GIoU represents the overlap between the predicted bounding box and the true bounding box; Btrue Denote the true bounding box of the annotation; B pred Denote the predicted bounding box; y c Denote the one-hot encoding of the class of the true bounding box; p c Denote the classification probability of the predicted bounding box; M c Denote the channel attention map output by the third activation function layer; ||M c ||1 denotes the L1-norm of M c ; M c [i, j] denotes the pixel at the i-th row and j-th column of M c ; M s Denote the spatial attention map output by the fourth activation function layer; ||M s ||1 denotes the L1-norm of M s ; M s [i, j] denotes the pixel at the i-th row and j-th column of M s ; H denotes the height of M c and M s ; W denotes the width of M c and M s ;

[0128] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, refer to the description in the method section.

[0129] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for aircraft positioning in remote sensing images based on a full-channel spatial attention module, characterized in that It includes the following steps: S1: Obtain the remote sensing image to be located; S2: After preprocessing the remote sensing image to be located, input it into the pre-trained remote sensing image aircraft positioning model to obtain the aircraft positioning result in the remote sensing image to be located; wherein, the remote sensing image aircraft positioning model includes a convolutional neural network, a full-channel spatial attention module, and a target detection and positioning module connected in sequence; The full-channel spatial attention module includes an adaptive average pooling layer, a first 1×1 convolutional layer, a first activation function layer, a second 1×1 convolutional layer, an adaptive max pooling layer, a third 1×1 convolutional layer, a second activation function layer, a fourth 1×1 convolutional layer, an element-wise addition unit, a third activation function layer, a first element-wise multiplication unit, a channel maximum value processing unit, a channel average value processing unit, a channel splicing layer, a 7×7 convolutional layer, a fourth activation function layer, and a second element-wise multiplication unit; The input ends of the adaptive average pooling layer and the adaptive max pooling layer are both connected to the output end of the convolutional neural network; The output end of the adaptive average pooling layer is connected to the input end of the element-wise addition unit through the first 1×1 convolutional layer, the first activation function layer, and the second 1×1 convolutional layer in sequence; The output end of the adaptive max pooling layer is connected to the input end of the element-wise addition unit through the third 1×1 convolutional layer, the second activation function layer, and the fourth 1×1 convolutional layer in sequence; The output end of the element-wise addition unit is connected to the input end of the first element-wise multiplication unit through the third activation function layer; The output end of the convolutional neural network is connected to the input end of the first element-wise multiplication unit; The output end of the first element-wise multiplication unit is respectively connected to the input ends of the channel maximum value processing unit and the channel average value processing unit; The output ends of the channel maximum value processing unit and the channel average value processing unit are both connected to the input end of the channel splicing layer; The output end of the channel splicing layer is connected to the input end of the second element-wise multiplication unit through the 7×7 convolutional layer and the fourth activation function layer in sequence; The output end of the first element-wise multiplication unit is connected to the input end of the second element-wise multiplication unit; The output end of the second element-wise multiplication unit is connected to the input end of the target detection and positioning module; Train the remote sensing image aircraft positioning model based on the following total loss function: L total = μL t + vL m + ηL s ; Among them, L total represents the total loss; L t , L m , L s successively represent the target loss, the feature enhancement consistency loss, and the attention sparsity loss; μ, v, η represent the weight coefficients; L t = αL box + βL cls ; L s = γ1L channel + γ2L spatial ; Among them, L box represents the bounding box regression loss; L cls represents the classification loss; α and β represent weight coefficients; X O represents the original features output by the convolutional neural network; X O [i] represents the i-th pixel of X O ; X en represents the enhanced features output by the full-channel spatial attention module; X en [i] represents the i-th pixel of X en ; N represents the number of pixels included in X O and X en ; L channel represents the channel attention sparsity loss; L spatial represents the spatial attention sparsity loss; γ1 and γ2 represent hyperparameters; ||X O [i] - X en [i]|| 2 represents the square of the Euclidean norm of X O [i] - X en [i]; L box = 1 - GIoU(B true , B pred ); L cls = -y c log(p c ) Among them, GIoU represents the overlap degree between the predicted bounding box and the ground truth bounding box; B true represents the annotated ground truth bounding box; B pred represents the predicted bounding box; y c represents the one-hot encoding of the class of the ground truth bounding box; p c represents the classification probability of the predicted bounding box; M c represents the channel attention map output by the third activation function layer; ||M c ||1 represents the L1-norm of M c ; M c [i, j] represents the pixel at the i-th row and j-th column of M c ; M s represents the spatial attention map output by the fourth activation function layer; ||M s ||1 represents the L1-norm of M s ; M s [i, j] represents the pixel at the i-th row and j-th column of M s ; H represents the height of M c and M s ; W represents the width of M c and M s ; 2. The remote sensing image aircraft positioning method based on the full-channel spatial attention module according to claim 1, characterized in that, The convolutional neural network includes one convolutional module / includes multiple convolutional modules connected in sequence; wherein, the convolutional module includes a convolutional layer, a non-linear activation function layer, a pooling layer, and a batch normalization layer connected in sequence.

3. A remote sensing image aircraft positioning method based on a full-channel spatial attention module according to claim 1, characterized in that, The target detection and positioning module includes a convolutional layer, a non-linear activation layer, a downsampling layer, and an upsampling layer connected in sequence; The input end of the convolutional layer of the target detection and positioning module is connected to the output end of the second element-wise multiplication unit; The output end of the upsampling layer is the output end of the remote sensing image aircraft positioning model.

4. A method for locating airplanes in remote sensing images based on a full-channel spatial attention module according to claim 1, characterized in that, The preprocessing in S2 includes scaling processing, normalization processing, and denoising processing.

5. The remote sensing image aircraft positioning method based on the full-channel spatial attention module according to claim 3, characterized in that, Obtain the pre-trained remote sensing image aircraft positioning model based on the following method: S21: Obtain an initial remote sensing image dataset; wherein, the initial remote sensing image dataset includes a number of remote sensing images containing airplanes; S22: Perform data annotation on the initial remote sensing image dataset; wherein, the annotation content includes the bounding box of the airplane, the coordinates of the bounding box, and the category of the bounding box; S23: Perform preprocessing on the annotated initial remote sensing image dataset to obtain a final remote sensing image dataset; S24: Input the final remote sensing image dataset into a remote sensing image airplane positioning model for training to obtain a pre-trained remote sensing image airplane positioning model.

6. The remote sensing image aircraft positioning method based on the full-channel spatial attention module according to claim 5, wherein, The preprocessing in S23 includes scaling processing, normalization processing, data augmentation processing, and denoising processing.

Images