Remote sensing panchromatic sharpening method and system based on cross-spectrum-space fusion network
By extracting feature information from multispectral and panchromatic remote sensing images through a cross-spectral-spatial fusion network, the problem of insufficient information fusion in existing methods is solved, and higher-precision, high-resolution multispectral remote sensing images are generated.
Patent Information
- Application Number
- CN202311230797.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-21
- Publication Date
- 2025-12-26
- Estimated Expiration
- 2043-09-21
AI Technical Summary
Existing deep learning methods cannot fully utilize the fitting ability of neural networks in panchromatic sharpening of remote sensing images, and cannot effectively fuse the spectral information of multispectral images and the spatial information of panchromatic images.
Design a cross-spectral-spatial fusion network. Extract spectral information of the multispectral branch and spatial information of the panchromatic branch through a cross-spectral-spatial attention module. Use a multi-level enhancement method to extract feature information and combine the basic residual module and attention module for feature fusion.
It generates higher-precision, high-resolution multispectral remote sensing images, which outperform existing methods in terms of subjective visual quality and objective evaluation metrics.
Smart Images
Figure CN117274093B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of remote sensing image processing, and relates to a remote sensing panchromatic sharpening method and system based on a cross-spectrum-space fusion network. BACKGROUND
[0002] High-resolution multispectral (HRMS) remote sensing images play a crucial role in geographic information management and are widely used in various fields, including change detection, scene classification, environmental pollution monitoring, and land surveying. However, due to the limitations of imaging devices, satellites usually capture low-resolution multispectral (LRMS) images and high-resolution panchromatic (PAN) images. Compared with MS images, PAN images have high spatial resolution and single-band characteristics. To obtain HRMS images, researchers use panchromatic sharpening techniques to fuse the spectral information of LRMS images with the spatial information of PAN images. Therefore, panchromatic sharpening algorithms are crucial fusion techniques in the practical application of remote sensing images.
[0003] With the significant progress of deep learning technology, remote sensing panchromatic sharpening methods have made significant progress based on deep learning methods. Masi et al. designed a remote sensing image fusion neural network consisting of only three convolutional layers. However, such a small number of layers may not fully utilize the excellent fitting ability of neural networks. Yang et al. used a deeper neural network to capture high-frequency details and integrate them into the up-sampled multispectral image. At the same time, Yuan et al. proposed a fusion network that combines multiscale and multi-depth convolutional neural networks. This method uses a multiscale feature extraction method to improve overall fusion performance. Shao et al. proposed a dual-branch convolutional neural network fusion method that combines residual learning, which can produce fusion results with small spectral distortion and rich spatial details. Liu et al. proposed a dual-flow fusion network that uses two branches to extract spatial information from LRMS images and PAN images, respectively, and then reconstructs the fusion image. Wang et al. proposed a novel dual-path fusion network that searches for the most similar image blocks between multispectral images and PAN images through global and local subnets and obtains high-frequency details.
[0004] In order to solve the problem that the above-mentioned deep learning method is insufficient in extracting spectral information from multispectral images and spatial information from panchromatic images, the application provides a method and system of a new cross-spectral spatial fusion network for panchromatic sharpening. In order to improve the extraction of spectral feature information from LRMS images and the extraction of spatial feature information from PAN images, a new cross-spectral spatial attention block is designed between two different feature extraction branches. The features extracted by the spectral attention module are injected into the feature representation to obtain a fused multi-level feature representation. SUMMARY
[0005] In view of the deficiencies of the existing panchromatic sharpening fusion technology of remote sensing images, the purpose of the application is to provide a new end-to-end panchromatic sharpening fusion scheme for remote sensing images.
[0006] The application provides a remote sensing panchromatic sharpening method based on a cross-spectral spatial fusion network, comprising the following steps:
[0007] Step a: down-sampling high-resolution multispectral remote sensing images to target low-resolution multispectral remote sensing images to construct a pair of high-low resolution multispectral remote sensing image sample library;
[0008] Step b: taking the low-resolution multispectral remote sensing images and the panchromatic remote sensing images as inputs, extracting feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images through a basic residual module;
[0009] Step c: inputting the extracted feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images into a cross-spectral spatial attention module, and using the cross-spectral spatial attention module to extract spectral information of a multispectral branch and spatial information of a panchromatic branch;
[0010] Step d: improving the fusion process and generating high-resolution multispectral images with detailed information by gradually enhancing the spectral and spatial feature representations of the remote sensing images.
[0011] Moreover, in step a, the high-resolution multispectral remote sensing images are down-sampled by a bicubic interpolation method to generate corresponding target low-resolution multispectral remote sensing images, and then the collected panchromatic remote sensing images are one-to-one corresponding to obtain a training sample library of high-low resolution multispectral remote sensing images and panchromatic remote sensing images.
[0012] Moreover, in step b, the low-resolution multispectral remote sensing images and the panchromatic remote sensing images in the training set are input into the basic residual module, and the feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images is extracted by the combination of the basic residual module.
[0013] Moreover, in step c, the extracted low-resolution multispectral remote sensing image and the feature information of the panchromatic remote sensing image are input into the cross-spectrum-space attention module, and the cross-spectrum-space attention module is combined using a spectrum attention module and a spatial attention module. Specifically, first, the panchromatic remote sensing image features extracted by the panchromatic branch are input into the spatial attention module to extract the spatial remote sensing information specific to the panchromatic image. Then, the multispectral feature information extracted by the multispectral branch is spliced, and the fused multispectral remote sensing image feature map is obtained through the spectrum attention module, and the fusion process is represented as:
[0014] M M-MS =F spe (F res (concat(F spa (M PAN↓2 ),M LRMS↑2 )))
[0015] Wherein F spe (), F spa () represents the spectrum attention module and the spatial attention module in the cross-spectrum-space attention module, F res () represents the basic residual module in the cross-spectrum-space attention module, Concat() represents the feature fusion layer, M M-MS represents the output of the intermediate cross-spectrum-space attention module, M PAN↓2 represents the extracted down-sampled 2 times panchromatic remote sensing image features, M LRMS↑2 represents the extracted up-sampled 2 times low-resolution multispectral remote sensing image features.
[0016] Moreover, the implementation of the three sub-module networks F res (), F spe (), F spe () in the cross-spectrum-space attention module is as follows,
[0017] In the sub-module network F res (), the feature information is sequentially input into a 3x3 convolution layer, a PReLU layer, a 3x3 convolution layer, and then the output is obtained through a point-by-point addition layer;
[0018] In the sub-module network F spa (), the feature map is subjected to a maximum pooling layer operation and an average pooling layer operation, respectively, and then the output is obtained through an activation function after fusion;
[0019] In the sub-module network F spe (), the feature map is sequentially input into an average pooling layer, a convolution layer, and an activation function layer, and then the output is obtained through a point-by-point addition layer.
[0020] Moreover, in step d, the feature map of the target remote sensing image is updated to fuse and generate a high-resolution multispectral remote sensing image result by further enhancing the spectral and spatial feature representation of the remote sensing image through the same module learning as in step c.
[0021] In another aspect, the present application also provides a remote sensing panchromatic sharpening fusion system based on a cross-spectrum-space fusion network, which is used to implement the remote sensing panchromatic sharpening method based on the cross-spectrum-space fusion network as described above.
[0022] Moreover, the system comprises the following modules,
[0023] The first module is to downsample the high-resolution multispectral remote sensing image to the target low-resolution multispectral remote sensing image to construct a pair of high-low resolution multispectral remote sensing image sample library;
[0024] The second module is to take the low-resolution multispectral remote sensing image and the panchromatic remote sensing image as input and extract the feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image through a basic residual module;
[0025] The third module is to input the extracted feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image into a cross-spectrum-space attention module to extract the spectral information of the multispectral branch and the spatial information of the panchromatic branch by using the cross-spectrum-space attention module;
[0026] The fourth module is to improve the fusion process and generate a high-resolution multispectral image with detailed information by gradually enhancing the spectral and spatial feature representation of the remote sensing image.
[0027] Alternatively, the system comprises a processor and a memory, the memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the remote sensing panchromatic sharpening method based on the cross-spectrum-space fusion network as described above.
[0028] Alternatively, the system comprises a readable storage medium, and the readable storage medium has a computer program stored thereon, and the computer program is executed to implement the remote sensing panchromatic sharpening method based on the cross-spectrum-space fusion network as described above.
[0029] The present application uses a cross-spectrum-space attention module to extract spectral information of a multispectral branch and spatial information of a panchromatic branch, and the difference from other inventions is that the present application designs a cross combination of a spatial attention module and a spectral attention module, and adopts a multi-level enhancement manner to extract feature information twice, so that the deep neural network can better fuse to generate a hyperspectral remote sensing image with higher resolution and more details. And the present application is different from other inventions in the use skill, the network of the present application ingeniously combines the spatial attention module and the spectral attention module, and efficiently extracts the feature information of the remote sensing image, so that the test cases of the present application are better than the existing most advanced remote sensing image panchromatic sharpening method in the subjective and objective results under the condition of shallow network depth.
[0030] The scheme proposed in the present application is better than other remote sensing image panchromatic sharpening methods in qualitative and quantitative evaluation, and can generate a higher precision high-resolution multispectral remote sensing image fusion result. BRIEF DESCRIPTION OF DRAWINGS
[0031] Figure 1 The present application is based on a cross-spectrum-space fusion network remote sensing panchromatic sharpening process schematic diagram.
[0032] Figure 2 The present application is based on a cross-spectrum-space fusion network remote sensing panchromatic sharpening network framework diagram.
[0033] Figure 3 The present application is a cross-spectrum-space attention module network architecture diagram.
[0034] Figure 4 The present application provides experimental results on public data sets, wherein (a) is a Bicubic image; (b) is an original high-resolution image; (c) is an experimental result of the present application. DETAILED DESCRIPTION
[0035] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.
[0036] REFERENCE Figures 1-3 The present application is a cross-spectrum-space fusion network remote sensing panchromatic sharpening method, which includes the following specific process steps:
[0037] Step a, down-sampling the high-resolution multispectral remote sensing image to the target low-resolution multispectral remote sensing image to construct a pair of high-low resolution multispectral remote sensing image sample library;
[0038] In the embodiment, it is further preferred to propose that the high-resolution multispectral remote sensing image is down-sampled by a bicubic interpolation method to generate a corresponding target low-resolution multispectral remote sensing image, and then the collected panchromatic remote sensing image is one-to-one corresponding to obtain a training sample library of high-low resolution multispectral remote sensing images and panchromatic remote sensing images.
[0039] Step b, taking the low-resolution multispectral remote sensing image and the panchromatic remote sensing image as input, extracting feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image through a basic residual module;
[0040] In the embodiment, it is further preferred to propose that the low-resolution multispectral remote sensing image and the panchromatic remote sensing image in the training set are input into the basic residual module, and it is preferably suggested that the feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image is extracted through the combination of multiple basic residual modules. The combination mode can be simple multi-layer superposition, or the combination of the obtained results after multiple superpositions and input, or a combination of the two modes, for example, in the embodiment, the panchromatic remote sensing image is input into the simple superposition of multiple basic residual modules, and the processing of the low-resolution multispectral remote sensing image input is a combination of the two modes.
[0041] Step c, inputting the extracted feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image into a cross-spectrum-space attention module, and using the cross-spectrum-space attention module to extract spectral information of the multispectral branch and spatial information of the panchromatic branch;
[0042] In the embodiment, it is further preferred to propose that the extracted feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image is first input into the cross-spectrum-space attention module, and the basic residual module, the spectral attention module and the spatial attention module are combined in the cross-spectrum-space attention module. Specifically, first, the panchromatic remote sensing image features extracted by the panchromatic branch are input into the spatial attention module to extract spatial remote sensing information specific to the panchromatic image. Then, the multispectral feature information extracted by the multispectral branch is spliced, input into the multi-layer basic residual module, and then input into the spectral attention module to obtain the fused multispectral remote sensing image feature map through the spectral attention module, and the fusion process is represented as:
[0043] M M-MS =F spe (F res (concat(F spa (M PAN↓2 ), M LRMS↑2 ))
[0044] where F spe () represents a spectral attention module and a spatial attention module within the cross-spectral-spatial attention module, F spa () represents a spectral attention module and a spatial attention module within the cross-spectral-spatial attention module, F res () represents a base residual module within the cross-spectral-spatial attention module, Concat() represents a feature fusion layer, M M-MS represents the output of the intermediate cross-spectral-spatial attention module, M PAN↓2 represents the extracted down-sampled 2 times panchromatic remote sensing image features, M LRMS↑2 represents the extracted up-sampled 2 times low-resolution multispectral remote sensing image features.
[0045] Furthermore, the three sub-module networks F res () within the cross-spectral-spatial attention module are implemented as follows, spa () within the cross-spectral-spatial attention module are implemented as follows, spe () within the cross-spectral-spatial attention module are implemented as follows,
[0046] The base residual module, i.e. the sub-module network F res (), sequentially inputs the feature information into a 3x3 convolution layer, a PReLU layer, a 3x3 convolution layer, and then obtains the output by point-wise addition of the obtained result and the input feature information.
[0047] The spectral attention module, i.e. the sub-module network F spa (), respectively performs a max-pooling layer operation and an average-pooling layer operation on the feature maps, and then obtains the output by activation function fusion.
[0048] The spatial attention module, i.e. the sub-module network F spe (), sequentially inputs the feature maps into an average-pooling layer, a convolution layer, and an activation function layer, and then obtains the output by point-wise addition of the obtained result and the input feature information.
[0049] Step d, by gradually enhancing the spectral and spatial feature representations of the remote sensing image, the fusion process is improved and a high-resolution multispectral image with detailed information is generated.
[0050] In step d of the embodiment, by taking the intermediate multispectral remote sensing image feature map extracted in step c above and the panchromatic remote sensing image as input, the same module learning as steps b and c is performed again to gradually enhance the spectral and spatial feature representations of the remote sensing image, and the feature map of the target remote sensing image is updated to generate a high-resolution multispectral remote sensing image result.
[0051] In specific implementation, the method provided by the technical scheme of the present application can be automatically run by a computer software technology, and the system device of the method, such as a computer readable storage medium storing the corresponding computer program of the technical scheme of the present application and a computer device including the running of the corresponding computer program, should also be within the protection scope of the present application.
[0052] In some possible embodiments, a remote sensing panchromatic sharpening system based on a cross-spectrum-space fusion network is provided, comprising the following modules,
[0053] A first module is configured to downsample high-resolution multi-spectral remote sensing images to target low-resolution multi-spectral remote sensing images to construct a high-low resolution multi-spectral remote sensing image sample library;
[0054] A second module is configured to take the low-resolution multi-spectral remote sensing images and the panchromatic remote sensing images as inputs, and extract feature information of the low-resolution multi-spectral remote sensing images and the panchromatic remote sensing images through a basic residual module;
[0055] A third module is configured to input the extracted feature information of the low-resolution multi-spectral remote sensing images and the panchromatic remote sensing images into a cross-spectrum-space attention module, and use the cross-spectrum-space attention module to extract spectral information of a multi-spectral branch and spatial information of a panchromatic branch;
[0056] A fourth module is configured to improve the fusion process and generate high-resolution multi-spectral images with detailed information by gradually enhancing the spectral and spatial feature representations of the remote sensing images.
[0057] In some possible embodiments, a remote sensing panchromatic sharpening system based on a cross-spectrum-space fusion network is provided, comprising a processor and a memory, the memory is configured to store program instructions, and the processor is configured to call the stored instructions in the memory to execute a remote sensing panchromatic sharpening method based on a cross-spectrum-space fusion network as described above.
[0058] In some possible embodiments, a remote sensing panchromatic sharpening system based on a cross-spectrum-space fusion network is provided, comprising a readable storage medium, and the readable storage medium has a computer program stored thereon, and the computer program is executed to implement a remote sensing panchromatic sharpening method based on a cross-spectrum-space fusion network as described above.
[0059] Reference Figure 3 , the application test experiment of the embodiment of the present application is as follows:
[0060] The experiment adopts public data sets WordView II and QuickBird data sets, selects 80 remote sensing images as a test set, each remote sensing image is 256*256*4 pixels in size, and a corresponding low resolution remote sensing image is obtained by downsampling four times, and the size of the low resolution remote sensing image is 64*64*4 pixels.
[0061] Compared with other panchromatic sharpening methods of remote sensing images, the effectiveness of the present application is expressed by providing experimental data, and the evaluation indexes of the comparison experiment results are shown in Table 1, and the experimental results are shown in Figure 1. Figure 4 (a) is a Bicubic image; (b) is an original high-resolution image; (c) is an experimental result of the present application.
[0062] Table 1
[0063] Method Bicubic GPPNN Band-Aware The invention PSNR 23.49 26.29 27.18 27.34 SSIM 0.6170 0.8704 0.8741 0.8810 SCC 0.8588 0.9720 0.9759 0.9766 Q 0.4481 0.8484 0.8504 0.8597
[0064] As can be seen from the above table experimental data, the method of the present application obtains a high score compared with the comparison method, that is, it is better than the comparison method.
[0065] In summary, the remote sensing panchromatic sharpening method and system based on the cross-spectrum-space fusion network provided by the present application first down-samples a high-resolution multispectral remote sensing image to a target low-resolution multispectral remote sensing image to construct a pair of high and low resolution multispectral remote sensing image sample library; then the low-resolution multispectral remote sensing image and the panchromatic remote sensing image are input, and the feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image is extracted through a basic residual module; the extracted feature information of the low-resolution multispectral remote sensing image and the panchromatic remote sensing image is input into a cross-spectrum-space attention module, and the cross-spectrum-space attention module is used to extract the spectral information of the multispectral branch and the spatial information of the panchromatic branch; the spectral and spatial feature representations of the remote sensing image are gradually enhanced, so as to improve the fusion process and generate a high-resolution multispectral image with detailed information. The network provided by the present application is better than other remote sensing panchromatic sharpening methods in qualitative and quantitative evaluation.
[0066] Compared with the prior art, the advantages and positive effects of the present application are as follows: the present application provides a new cross-spectrum-space fusion network for panchromatic sharpening of remote sensing images, which mainly adopts a double-branch feature extraction method and a progressive learning method to fuse multi-level features to obtain a final fusion image; meanwhile, a new cross-spectrum-space attention module is designed, which can better extract the spectral feature information of the low-resolution multispectral remote sensing image and the spatial feature information of the panchromatic remote sensing image. The superiority of the method proposed in this paper is effectively verified from the subjective visual quality and objective evaluation indexes on the public data set.
[0067] The specific embodiments described herein are merely illustrative of the spirit of the application. Various modifications or changes in the specific embodiments described herein can occur to those skilled in the art to which the application pertains without departing from the spirit of the application, and it is understood that such modifications or changes are to be considered as within the scope of the application as defined by the appended claims.
Claims
1. A method for remote sensing panchromatic sharpening based on a cross-spectrum-space fusion network, characterized in that, The method comprises the following steps: Step a: down-sampling high-resolution multispectral remote sensing images to target low-resolution multispectral remote sensing images to construct a high-low resolution multispectral remote sensing image sample library; Step b: taking low-resolution multispectral remote sensing images and panchromatic remote sensing images as inputs, extracting feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images through a basic residual module; Step c: inputting the extracted feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images into a cross-spectrum-space attention module, and extracting spectral information of a multispectral branch and spatial information of a panchromatic branch by using the cross-spectrum-space attention module; The cross-spectrum-space attention module is combined by using a spectral attention module and a spatial attention module, which comprises the following steps: first, inputting panchromatic remote sensing image features extracted by the panchromatic branch into the spatial attention module to extract spatial remote sensing information specific to the panchromatic image; then, splicing multispectral feature information extracted by the multispectral branch to obtain a fused multispectral remote sensing image feature map through the spectral attention module, and the fusion process is represented as: wherein , denotes a spectral attention module and a spatial attention module within the cross-spectral-spatial attention module, denotes a base residual module within the cross-spectral-spatial attention module, denotes a feature fusion layer, denotes an output of the intermediate cross-spectral-spatial attention module, denotes extracted down-sampled by a factor of 2 panchromatic remote sensing image features, denotes extracted up-sampled by a factor of 2 low-resolution multispectral remote sensing image features; Step d: improving the fusion process by gradually enhancing spectral and spatial feature representations of remote sensing images, thereby generating high-resolution multispectral images with detailed information.
2. The remote sensing panchromatic sharpening method based on the cross-spectrum-space fusion network according to claim 1, characterized in that: In step a, the high-resolution multispectral remote sensing images are down-sampled by a bicubic interpolation method to generate corresponding target low-resolution multispectral remote sensing images, and then the collected panchromatic remote sensing images are corresponded one by one to obtain a training sample library of high-low resolution multispectral remote sensing images and panchromatic remote sensing images.
3. The remote sensing panchromatic sharpening method based on the cross-spectrum-space fusion network according to claim 1, characterized in that: In step b, the low-resolution multispectral remote sensing images and the panchromatic remote sensing images are taken as inputs, and the feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images is extracted through the basic residual module.
4. The remote sensing panchromatic sharpening method based on the cross-spectrum-space fusion network according to claim 1, characterized in that: Three sub-module networks within a cross-spectrum-spatial attention module , , Implementations of the above are as follows, Submodule network In the embodiment, the feature information is sequentially input to convolutional layer, PReLU layer, convolutional layer, and then output is obtained through a point-wise addition layer. Submodule network In the embodiment, the feature map is respectively subjected to a max pooling layer operation and an average pooling layer operation, and then fused to obtain an output after an activation function. Submodule network In the embodiment, the feature map is sequentially input to an average pooling layer, a convolution layer, an activation function layer, and then an element-wise addition layer to obtain the output.
5. The remote sensing panchromatic sharpening method based on the cross-spectrum-space fusion network according to claim 1 or 2 or 3 or 4, characterized in that: In step d, the intermediate multispectral remote sensing image feature map extracted in step c is inputted into the same module in step c again to gradually enhance the spectral and spatial feature representations of the remote sensing images, thereby updating the feature map of the target remote sensing image to generate a high-resolution multispectral remote sensing image result.
6. A remote sensing panchromatic sharpening system based on a cross-spectrum-space fusion network, characterized in that: The application discloses a cross-spectrum-space fusion network-based remote sensing panchromatic sharpening method.
7. The remote sensing panchromatic sharpening system based on the cross-spectrum-space fusion network according to claim 6, characterized in that: The method comprises the following modules, A first module is configured to down-sample high-resolution multispectral remote sensing images to target low-resolution multispectral remote sensing images to construct a high-low resolution multispectral remote sensing image sample library; A second module is configured to take low-resolution multispectral remote sensing images and panchromatic remote sensing images as inputs, and extract feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images through a basic residual module; A third module is configured to input the extracted feature information of the low-resolution multispectral remote sensing images and the panchromatic remote sensing images into a cross-spectrum-space attention module, and extract spectral information of a multispectral branch and spatial information of a panchromatic branch by using the cross-spectrum-space attention module; A fourth module is configured to improve a fusion process by gradually enhancing spectral and spatial feature representations of remote sensing images, thereby generating high-resolution multispectral remote sensing images with detailed information.
8. A remote sensing panchromatic sharpening device based on a cross-spectrum-space fusion network, characterized in that: The application relates to a remote sensing panchromatic sharpening method based on a cross-spectrum-space fusion network, and comprises a processor and a memory.
9. A readable storage medium characterized by: The application relates to a remote sensing panchromatic sharpening method based on a cross-spectrum-space fusion network, and comprises a processor and a memory.