Unmanned aerial vehicle remote sensing image defogging method, system and device based on UAVD-Net and medium

Abstract

The invention discloses a UAVD-Net-based unmanned aerial vehicle remote sensing image defogging method, system and device and a medium, and the method comprises the steps: constructing a non-uniform haze remote sensing image data set which comprises a training set, a verification set and a test set; an unmanned aerial vehicle remote sensing image defogging network based on UAVD-Net is constructed; training an unmanned aerial vehicle remote sensing image defogging network based on UAVD-Net by using the training set and the verification set to obtain an optimal training weight file Dehaze.pt; performing unmanned aerial vehicle remote sensing image defogging on the unmanned aerial vehicle remote sensing image defogging network based on UAVD-Net by using the test set and the optimal training weight file Dehaze.pt to obtain an unmanned aerial vehicle remote sensing image defogging result; the system, the equipment and the medium are used for implementing the method. The multi-level global information capture module, the self-adaptive local information enhancement module and the cross channel feature fusion module are introduced into the network, the global information and the local features of the image are effectively fused, complementation is achieved, the defogging precision of the unmanned aerial vehicle remote sensing image under the non-uniform haze condition is effectively improved, and the defogging accuracy of the unmanned aerial vehicle remote sensing image is improved. The method has the advantages of high adaptability and high defogging precision.

Description

Technical Field

[0001] The present invention belongs to the field of image processing technology, and further relates to deep learning and digital image processing technology, and specifically is a method, system, device and medium for defogging unmanned aerial vehicle remote sensing images based on UAVD-Net. The present invention can be used for defogging unmanned aerial vehicle optical remote sensing images. Background Art

[0002] As a flexible and efficient aerial remote sensing tool, unmanned aerial vehicles (UAVs) have played an important role in environmental monitoring, disaster assessment, urban planning, agricultural management and other fields. Especially in complex terrain or extreme climatic conditions, UAVs can obtain high-resolution image data and provide key support for decision-making. However, due to the presence of haze, dust and other particles in the atmosphere, remote sensing images are often disturbed during the acquisition process, resulting in reduced image contrast and blurred details, affecting the reliability and accuracy of the data. Image defogging technology has therefore become an important topic in UAV remote sensing applications.

[0003] Traditional image dehazing methods are mostly based on theories and technologies in the fields of image enhancement and image restoration. However, in practical applications, especially when images taken by drones are often faced with challenges such as complex lighting, changing scenes, and uneven haze levels, the performance of existing methods is limited. With the rapid development of deep learning and computer vision technology, dehazing methods based on convolutional neural networks (CNNs) and transformers have gradually become a research hotspot, and they have shown great potential in extracting image features and improving detail restoration.

[0004] The patent application document with publication number CN118052737A discloses a perception-guided UAV image defogging algorithm based on superpixel scene prior. The implementation steps of the method are: introducing superpixel scene prior to reduce the calculation of the UAV defogging process, the image defogging process can be converted from RGB color space to Lab color space to avoid confusion and redundant information between different chromaticities, so as to select reliable areas for efficient defogging through the L channel. Considering the influence of light inhomogeneity in the defogging process, the invention designs a guided filtering algorithm based on simple linear iterative clustering, which replaces a large number of guided windows with superpixel clustering windows with similar color blocks while retaining complementary information. In order to improve the perception ability after defogging, the superpixel segmentation and target detection results are quantitatively analyzed, and the alternating direction multiplier method is used to design a collaborative optimization feedback iteration mechanism of perception and defogging to enhance the effectiveness of UAV visual tasks in foggy environments. However, the invention still has the disadvantage that the method depends on the atmospheric light intensity value, and the defogging effect of the method is poor when there is a large area of ​​blank area in the image.

[0005] The patent application document with publication number CN118314053A discloses a method for defogging drone inspection aerial images based on a neural network. The implementation steps of the method are: constructing an image data set based on drone inspection aerial images; constructing a heuristic perception defogging neural network, training the heuristic perception defogging neural network based on the image data set to obtain a trained heuristic perception defogging neural network; inputting the drone inspection aerial images to be defogged into the trained heuristic perception defogging neural network for processing to obtain a fog-free and clear aerial image. However, the invention still has the disadvantage that the method can only ignore local details in the image, resulting in blurring or detail loss in local areas of the defogged image.

[0006] The defects and shortcomings of the above-mentioned prior art are summarized as follows:

[0007] 1. Many existing methods assume that haze is uniformly distributed and use simple models to estimate the impact of haze. However, in real scenes, haze often presents non-uniformity, which can lead to a decrease in the quality of dehazing results, especially in images with high dynamic range or complex terrain.

[0008] 2. When dealing with global haze, existing dehazing algorithms tend to ignore local details in the image, resulting in blurring or detail loss in local areas of the dehazed image. Summary of the invention

[0009] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a UAVD-Net-based UAV remote sensing image defogging method, system, device and medium, by designing a multi-level global information capture module (MGIC), extracting and fusion of global feature layers layer by layer to enhance the model's ability to understand complex scenes and improve the accuracy of defogging; by designing an adaptive local information enhancement module (ALIE), the texture detail information in the image can be effectively acquired and enhanced to improve the defogging accuracy; a cross-channel feature fusion module (CFF) is designed to fuse global and local information through a cross-channel mechanism to maintain the overall structure of the image and enhance the clarity of local details, thereby obtaining a natural and clear defogging image; this method is used to solve the problem that the existing UAV defogging method has low defogging quality in non-uniform haze conditions, and the defogging image is blurred or has detail loss in local areas.

[0010] In order to achieve the above object, the technical solution adopted by the present invention is:

[0011] A method for defogging UAV remote sensing images based on UAVD-Net, comprising the following steps:

[0012] Step 1: Construct a non-uniform haze remote sensing image dataset, including a training set, a validation set, and a test set;

[0013] Step 2: Build a UAV remote sensing image defogging network based on UAVD-Net, including an encoder, a texture detail feature extraction subnetwork, and a decoder.

[0014] Step 3: Use the training set and validation set in the non-uniform haze remote sensing image dataset constructed in step 1 to train the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2 to obtain the optimal training weight file Dehaze.pt;

[0015] Step 4: Use the test set in the non-uniform haze remote sensing image dataset constructed in step 1 and the optimal training weight file Dehaze.pt obtained in step 3 to perform UAV remote sensing image dehazing on the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2 to obtain the UAV remote sensing image dehazing results.

[0016] The UAVD-Net-based UAV remote sensing image defogging network in step 2 includes an encoder, a texture detail feature extraction subnetwork and a decoder;

[0017] The encoder includes a CNN network layer, five multi-level global information capture modules (MGIC) and a summation layer; the multi-level global information capture module (MGIC) includes a patch embedding (PEb) layer, a positional encoding layer (Positional Encoding, PE), a multi-level encoder (Multi-Level Encoder, MLE) and a dehazing enhancement layer (Dehazing Enhancement Layer, DEL), wherein the multi-level encoder includes three multi-head self-attention mechanism layers (Multi-head Self-Attention Mechanism, MHSA), three feed forward network layers (Feed Forward Network, FFN), two normalization layers (Batch Normalization, BN) and a linear normalization layer (Linear Normalization, LN); the dehazing enhancement layer includes a convolution layer (Conv), a ReLU activation function, three hybrid attention mechanism layers (Convolution Block Attention Module, CBAM) and three normalization (BN) layers; the multi-level global information capture module (MGIC) is expressed as follows:

[0018] Foutput =DEL(MLE(PE+PEb(F input )));

[0019] The texture detail feature extraction subnetwork (Texture Detail Feature Extraction) includes a Conv layer and four adaptive local information enhancement modules (ALIE); the adaptive local information enhancement module (ALIE) includes two serially connected Conv layers, two multi-head self-attention mechanism layers (Multi-head Self-Attention Mechanism, MHSA), two summation and normalization layers (Add&Norm, A&N) and a position encoding layer (Positional Encoding, PE); the adaptive local information enhancement module (ALIE) is expressed as follows:

[0020] F ALIE =F3+Conv(Conv(F'))

[0021] F3=A&N(MHSA(F2),F2)

[0022] F2=A&N(MHSA(F1))

[0023] F1 = Conv(Conv(F')) + PE;

[0024] The decoder (Decoder) includes a cross-channel feature fusion module (CFF); the cross-channel feature fusion module (CFF) fuses global features and local features, including an upsampling layer (Up), a channel attention mechanism (CAM), two Conv layers, a ReLU activation function, two summing layers (Cat), a Conv layer with a convolution kernel of 3, a Conv layer with a convolution kernel of 5, and a Conv layer with a convolution kernel of 7; the cross-channel feature fusion module (CFF) is expressed as follows:

[0025] F * =Conv(Cat(Conv(Cat(ReLU(Conv(up(F G ),CAM(F L )),up(F)))) i )).

[0027] In step 3, the number of training rounds is set to be greater than or equal to 150, and the batch size is set to be greater than or equal to 32.

[0028] In step 4, the test batch size is set to be greater than or equal to 8.

[0029] The present invention also provides a UAV remote sensing image defogging system based on UAVD-Net, comprising:

[0030] Data acquisition module: used to construct non-uniform haze remote sensing image datasets, including training sets, validation sets, and test sets;

[0031] Network construction module: used to build a UAVD-Net-based UAV remote sensing image defogging network, including an encoder, a texture detail feature extraction subnetwork, and a decoder.

[0032] Network training module: used to train the UAVD-Net-based UAV remote sensing image dehazing network using the training set and validation set in the non-uniform haze remote sensing image dataset to obtain the optimal training weight file Dehaze.pt;

[0033] Dehazing module: It is used to use the test set in the non-uniform haze remote sensing image dataset and the optimal training weight file Dehaze.pt to perform UAV remote sensing image dehazing on the UAVD-Net-based UAV remote sensing image dehazing network to obtain the UAV remote sensing image dehazing results.

[0034] The present invention also provides a UAVD-Net-based UAV remote sensing image defogging device, comprising:

[0035] Memory: a computer program for storing the above-mentioned UAVD-Net-based UAV remote sensing image defogging method, which is a computer-readable device;

[0036] Processor: used to implement the UAVD-Net-based UAV remote sensing image defogging method when executing the computer program.

[0037] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it can implement the UAVD-Net-based UAV remote sensing image defogging method.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention designs five multi-level global information capture modules (MGIC) in the encoder, which extract and fuse global feature layers layer by layer to enhance the model's ability to understand complex scenes and improve the accuracy of non-uniform haze defogging.

[0040] 2. The present invention designs four adaptive local information enhancement modules (ALIE) in the texture detail feature extraction network, which can effectively obtain and enhance the texture detail information in the image and improve the defogging accuracy.

[0041] 3. The present invention designs a cross-channel feature fusion module (CFF) in the decoder, which fuses global and local information through a cross-channel mechanism to maintain the overall structure of the image and enhance the clarity of local details, thereby obtaining a natural and clear dehazed image.

[0042] In summary, the present invention designs a brand-new network structure, which effectively integrates the global information and local features of the image by introducing a multi-level global information capture MGIC module, an adaptive local information enhancement module (ALIE) and a cross-channel feature fusion module CFF into the network, and achieves complementarity. It can effectively improve the defogging accuracy of UAV remote sensing images under non-uniform haze conditions, and has the advantages of strong adaptability and high defogging accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the principle flow of an embodiment of the present invention.

[0044] Figure 2 It is a network structure diagram of UAV remote sensing image defogging based on UAVD-Net according to an embodiment of the present invention.

[0045] Figure 3 It is a structural diagram of the MGIC module of an embodiment of the present invention.

[0046] Figure 4 It is a structural diagram of the ALIE module of an embodiment of the present invention.

[0047] Figure 5 4 is a structural diagram of a CFF module according to an embodiment of the present invention.

[0048] Figure 6 This is a diagram of the defogging accuracy of UAV remote sensing images of the present invention and the existing method. DETAILED DESCRIPTION

[0049] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0050] like Figure 1 As shown, a UAV remote sensing image defogging method based on UAVD-Net includes the following steps:

[0051] Step 1: Construct a non-uniform haze remote sensing image dataset, including a training set, a validation set, and a test set;

[0052] Uniform haze images are selected from the public UAV foggy remote sensing image dataset to form a non-uniform haze remote sensing image dataset, and the non-uniform haze remote sensing image dataset is divided into training set, validation set and test set according to the proportion.

[0053] like Figure 2 As shown, step 2: construct a UAV remote sensing image defogging network based on UAVD-Net, including an encoder, a texture detail feature extraction subnetwork (Texture Detail Feature Extraction) and a decoder;

[0054] like Figure 3 As shown, step 201: the encoder includes a CNN network layer, five multi-level global information capture modules (MGIC) and a summation layer. The multi-level global information capture module (MGIC) includes a patch embedding (PEb) layer, a positional encoding layer (Positional Encoding, PE), a multi-level encoder (Multi-Level Encoder, MLE) and a dehazing enhancement layer (Dehazing Enhancement Layer, DEL), wherein the multi-level encoder includes three multi-head self-attention mechanism layers (Multi-head Self-Attention Mechanism, MHSA), three feed forward network layers (Feed Forward Network, FFN), two normalization layers (Batch Normalization, BN) and a linear normalization layer (Linear Normalization, LN); the dehazing enhancement layer includes a convolution layer (Conv), a ReLU activation function, three hybrid attention mechanism layers (Convolution Block Attention Module, CBAM) and three normalization (BN) layers; the multi-level global information capture module (MGIC) is expressed as follows:

[0055] F output =DEL(MLE(PE+PEb(F input )));

[0056] like Figure 4As shown, step 202: the texture detail feature extraction subnetwork (Texture Detail Feature Extraction) includes a Conv layer and four adaptive local information enhancement modules (ALIE); the adaptive local information enhancement module (ALIE) includes two serially connected Conv layers, two multi-head self-attention mechanism layers (Multi-head Self-Attention Mechanism, MHSA), two summation and normalization layers (Add&Norm, A&N) and a position encoding layer (Positional Encoding, PE); the adaptive local information enhancement module (ALIE) is expressed as follows:

[0057] F ALIE =F3+Conv(Conv(F'))

[0058] F3=A&N(MHSA(F2),F2)

[0059] F2=A&N(MHSA(F1))

[0060] F1 = Conv(Conv(F')) + PE;

[0061] like Figure 5 As shown, step 203: the decoder includes a cross-channel feature fusion module (CFF); the cross-channel feature fusion module (CFF) fuses global features and local features, including an upsampling layer (up), a channel attention mechanism (CAM), two Conv layers, a ReLU activation function, two summing layers (Cat), a Conv layer with a convolution kernel of 3, a Conv layer with a convolution kernel of 5, and a Conv layer with a convolution kernel of 7. The cross-channel feature fusion module (CFF) is expressed as follows:

[0062] F * =Conv(Cat(Conv(Cat(ReLU(Conv(up(F G ),CAM(F L )),up(F)))) i )).

[0064] Step 3: Set the number of training rounds to be greater than or equal to 150 and the batch size to be greater than or equal to 32. Use the training set and validation set in the non-uniform haze remote sensing image dataset constructed in step 1 to train the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2 to obtain the optimal training weight file Dehaze.pt.

[0065] Step 4: Set the test batch size to be greater than or equal to 8, use the test set in the non-uniform haze remote sensing image dataset constructed in step 1 and the optimal training weight file Dehaze.pt obtained in step 3 to perform UAV remote sensing image dehazing on the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2, and obtain the UAV remote sensing image dehazing results.

[0066] like Figure 2 As shown, the present invention also provides a UAV remote sensing image defogging system based on UAVD-Net, comprising:

[0067] Data acquisition module: used to implement the construction of non-uniform haze remote sensing image dataset in step 1, including training set, validation set and test set;

[0068] Network construction module: used to implement the construction of the UAVD-Net-based UAV remote sensing image defogging network in step 2, including encoder, texture detail feature extraction subnetwork (Texture Detail Feature Extraction) and decoder;

[0069] Network training module: used to implement the training set and validation set in the non-uniform haze remote sensing image dataset constructed in step 1 in step 3 to train the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2, and obtain the optimal training weight file Dehaze.pt;

[0070] Dehazing module: It is used to implement the test set in the non-uniform haze remote sensing image dataset constructed in step 1 and the optimal training weight file Dehaze.pt obtained in step 3 in step 4 to perform UAV remote sensing image dehazing on the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2, and obtain the UAV remote sensing image dehazing result.

[0071] The present invention also provides a UAVD-Net-based UAV remote sensing image defogging device, comprising:

[0072] Memory: a computer program for storing the above-mentioned UAVD-Net-based UAV remote sensing image defogging method, which is a computer-readable device;

[0073] Processor: used to implement the UAVD-Net-based UAV remote sensing image defogging method when executing the computer program.

[0074] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it can implement the UAVD-Net-based UAV remote sensing image defogging method.

[0075] The effect of the present invention is further described below in conjunction with simulation experiments.

[0076] 1. Simulation experiment conditions

[0077] The hardware platform of the simulation experiment of the present invention is: the processor is Intel i9-13900K, the main frequency is 3.0GHz, and the running memory is 64G.

[0078] The software platform of the simulation experiment platform of the present invention is: Windows 11 operating system and PyCharm, PyTorch2.1.0, and CUDA 12.1.

[0079] 2. Simulation steps

[0080] The training set and validation set of the public non-uniform haze remote sensing image dataset constructed in step 1 are input into the UAVD-Net-based UAV remote sensing image defogging network for optimization training. All remote sensing image defogging uses single-scale training, the image input size is 512×512 pixels, and the number of epochs is set to 150.

[0081] 3. Simulation content and results analysis

[0082] The simulation experiment of the present invention performs defogging on three UAV remote sensing images containing non-uniform haze, and the results are averaged. Figure 6 shown.

[0083] Combine the following Figure 6 The simulation effect of the present invention is further described.

[0084] Figure 6 This is a comparison chart of the dehazing accuracy of the present invention and the existing method. The comparison algorithm is DehazeFormer (Song, Y., He, Z., Qian, H., & Du, X. [J] (2022). Vision Transformers for Single Image Dehazing. IEEE Transactions on Image Processing, 32, 1927-1941.), and the test image size is 512×512The accuracy evaluation indicators are peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). PSNR is an indicator used to measure the quality of images or videos. The higher the value, the better the image quality and the smaller the noise. SSIM evaluates the similarity of two images by comparing brightness, contrast, and structure. The SSIM value is between 0 and 1. The closer the value is to 1, the more similar the two images are.

[0085] like Figure 6 As shown, it can be seen that the PSNR and SSIM of the present invention are higher than those of DehazeFormer. Experimental results show that the multi-level global information capture module (MGIC) designed by the present invention extracts and fuses the global feature layers layer by layer to enhance the model's ability to understand complex scenes and improve the accuracy of non-uniform haze dehazing. At the same time, the adaptive local information enhancement module (ALIE) can effectively acquire and enhance the texture detail information in the image and improve the dehazing accuracy. In addition, the cross-channel feature fusion module (CFF) fuses global and local information through a cross-channel mechanism to maintain the overall structure of the image and enhance the clarity of local details, thereby obtaining a natural and clear dehazed image. In general, the UAVD-Net-based UAV remote sensing image dehazing method proposed in the present invention significantly improves the dehazing accuracy of remote sensing images and meets the requirements of high-precision dehazing.

Claims

1. A UAV remote sensing image defogging method based on UAVD-Net, characterized in that: The following steps are involved: Step 1: Construct a non-uniform haze remote sensing image dataset, including a training set, a validation set, and a test set; Step 2: Build a UAV remote sensing image defogging network based on UAVD-Net, including an encoder, a texture detail feature extraction subnetwork, and a decoder. Step 3: Use the training set and validation set in the non-uniform haze remote sensing image dataset constructed in step 1 to train the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2 to obtain the optimal training weight file Dehaze.pt; Step 4: Use the test set in the non-uniform haze remote sensing image dataset constructed in step 1 and the optimal training weight file Dehaze.pt obtained in step 3 to perform UAV remote sensing image dehazing on the UAVD-Net-based UAV remote sensing image dehazing network constructed in step 2 to obtain the UAV remote sensing image dehazing results.

2. The method for defogging UAV remote sensing images based on UAVD-Net according to claim 1, characterized in that: The UAVD-Net-based UAV remote sensing image defogging network in step 2 includes an encoder, a texture detail feature extraction subnetwork and a decoder; The encoder includes a CNN network layer, five multi-level global information capture modules (MGIC) and a summation layer; The multi-level global information capture module (MGIC) includes a patch embedding (PEb) layer, a positional encoding layer (PE), a multi-level encoder (MLE) and a dehazing enhancement layer (DEL), wherein the multi-level encoder includes three multi-head self-attention mechanism layers (MHSA), three feed forward network layers (FFN), two normalization layers (Batch Normalization, BN) and a linear normalization layer (LN); the dehazing enhancement layer includes a convolution layer (Conv), a ReLU activation function, three hybrid attention mechanism layers (ConvolutionBlockAttention Module, CBAM) and three normalization (BN) layers; the multi-level global information capture module (MGIC) is expressed as follows: F output =DEL(MLE(PE+PEb(F input ))); The texture detail feature extraction subnetwork (TextureDetailFeatureExtraction) includes a Conv layer and four adaptive local information enhancement modules (ALIE); the adaptive local information enhancement module (ALIE) includes two serially connected Conv layers, two multi-head self-attention mechanism layers (Multi-headSelf-AttentionMechanism, MHSA), two summation and normalization layers (Add&Norm, A&N) and a position encoding layer (PositionalEncoding, PE); the adaptive local information enhancement module (ALIE) is expressed as follows: F ALIE =F3+Conv(Conv(F')) F3=A&N(MHSA(F2),F2) F2=A&N(MHSA(F1)) F1 = Conv(Conv(F')) + PE; The decoder (Decoder) includes a cross-channel feature fusion module (CFF); the cross-channel feature fusion module (CFF) fuses global features and local features, including an upsampling layer (Up), a channel attention mechanism (CAM), two Conv layers, a ReLU activation function, two summing layers (Cat), a Conv layer with a convolution kernel of 3, a Conv layer with a convolution kernel of 5, and a Conv layer with a convolution kernel of 7; the cross-channel feature fusion module (CFF) is expressed as follows: F * =Conv(Cat(Conv(Cat(ReLU(Conv(up(F G ),CAM(F L )),up(F)))) i ))。 3. The method for defogging UAV remote sensing images based on UAVD-Net according to claim 1, characterized in that: In step 3, the number of training rounds is set to be greater than or equal to 150, and the batch size is set to be greater than or equal to 32.

4. The method for defogging an unmanned aerial vehicle remote sensing image based on UAVD-Net according to claim 1, characterized in that: In step 4, the test batch size is set to be greater than or equal to 8.

5. A UAVD-Net-based UAV remote sensing image defogging system based on the method according to any one of claims 1 to 4, characterized in that: include: Data acquisition module: used to construct non-uniform haze remote sensing image datasets, including training sets, validation sets, and test sets; Network construction module: used to build a UAVD-Net-based UAV remote sensing image defogging network, including an encoder, a texture detail feature extraction subnetwork (Texture Detail Feature Extraction) and a decoder (Decoder); Network training module: used to train the UAVD-Net-based UAV remote sensing image dehazing network using the training set and validation set in the non-uniform haze remote sensing image dataset to obtain the optimal training weight file Dehaze.pt; Dehazing module: It is used to use the test set in the non-uniform haze remote sensing image dataset and the optimal training weight file Dehaze.pt to perform UAV remote sensing image dehazing on the UAVD-Net-based UAV remote sensing image dehazing network to obtain the UAV remote sensing image dehazing results.

6. A UAV remote sensing image defogging device based on UAVD-Net, characterized in that: include: Memory: a computer program storing a method for defogging an unmanned aerial vehicle remote sensing image based on UAVD-Net as described in any one of claims 1 to 4, which is a computer-readable device; Processor: used to implement the UAVD-Net-based UAV remote sensing image defogging method described in any one of claims 1 to 4 when executing the computer program.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it can implement the UAVD-Net-based UAV remote sensing image defogging method described in any one of claims 1 to 4.

Images