A mangrove optical remote sensing image super-resolution reconstruction method, device, equipment, medium and product

Abstract

The present application relates to the field of image processing, and proposes a mangrove optical remote sensing image super-resolution reconstruction method, device, equipment, medium and product. The present application adopts an improved residual network structure, designs a new residual module containing switchable hollow convolution, activation function and linear scaling, and is used for enhancing and extracting deep semantic features of mangroves; a new feature fusion module containing a pixel attention mechanism is constructed, the semantic features of the mangroves are adaptively weighted, and the effective feature expression is enhanced; finally, the fused features and the original low-resolution image are added and spliced to output a high-resolution remote sensing super-resolution image. The present application can effectively improve the spatial resolution of the remote sensing image and restore the texture details, improve the image quality, and further greatly improve the accuracy of the mangrove recognition task.

Description

Technical Field

[0001] This application relates to the field of image processing, and in particular to a method, apparatus, equipment, medium, and product for super-resolution reconstruction of mangrove optical remote sensing images. Background Technology

[0002] Mangroves are vital marine ecosystems globally, playing crucial roles in water purification and coastal erosion prevention. However, in recent decades, due to human and natural factors, global mangrove forests have rapidly declined, fragmenting into small patches and gradually disappearing. To address this mangrove loss, the Global Mangrove Consortium, in collaboration with member institutions and governments, has developed a high-resolution Global Mangrove Watch (GMW) map to identify the most important protected areas and establish a long-term change tracking system to systematically record the annual and multi-year evolution of mangrove cover. Therefore, long-term dynamic monitoring of mangroves is crucial for revealing patterns of cover change, tracing the causes of loss, developing scientific and long-term protection and restoration strategies, and evaluating the long-term effectiveness of protection measures.

[0003] Satellite optical remote sensing technology, with its excellent spatiotemporal coverage and multispectral information, is widely used in mangrove monitoring and evaluation. Low-to-medium resolution remote sensing images, represented by the Landsat series, play a crucial role in long-term mangrove monitoring. However, when extracting mangrove distribution information, problems exist such as insufficient spatial resolution, inaccurate mangrove patch boundary delineation, and the easy omission of small patches and sparse mangroves. A feasible solution is to utilize super-resolution reconstruction technology to perform image super-resolution reconstruction tasks using satellite optical remote sensing images. Although convolutional neural network-based methods have achieved excellent performance in super-resolution tasks on natural images, the variability of remote sensing images means that super-resolution models applicable to natural images may not be suitable for remote sensing images. Furthermore, traditional convolutional neural network (CNN) models are insufficient in capturing the spectral and textural features of mangroves in remote sensing images. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, device, medium, and product for super-resolution reconstruction of mangrove optical remote sensing images, so as to improve the quality of super-resolution reconstruction of mangrove optical remote sensing images.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for super-resolution reconstruction of mangrove optical remote sensing images, including: Acquire optical remote sensing images of mangroves and perform data preprocessing to construct a super-resolution reconstruction simulation dataset; The super-resolution reconstruction simulation dataset is input into the super-resolution network, which includes a feature extraction module, a feature fusion module, and an image generation module. The shallow features and deep semantic features of the super-resolution reconstruction simulation dataset are obtained through the feature extraction module. The feature fusion module adaptively weights and fuses the shallow features and deep semantic features to obtain an effective feature representation. The image generation module generates and stitches together super-resolution images based on the effective feature representation and the super-resolution reconstruction simulation dataset to obtain spatially continuous, band-complete, and high-resolution remote sensing super-resolution images that meet the application standards of remote sensing images.

[0006] Secondly, this application provides a mangrove optical remote sensing image super-resolution reconstruction device, including a dataset construction unit, a feature extraction unit, a feature fusion unit, and an image generation unit; the dataset construction unit, the feature extraction unit, the feature fusion unit, and the image generation unit are sequentially and communicatively connected to collaboratively complete the mangrove super-resolution image reconstruction task, with the following specific functions: The dataset construction unit is used to acquire optical remote sensing images of mangroves and perform data preprocessing to construct a standardized super-resolution reconstruction simulation dataset adapted to the input of the super-resolution network. The feature extraction unit is used to obtain the shallow features and deep semantic features of the super-resolution reconstruction simulation dataset; The feature fusion unit is used to adaptively weight and fuse the shallow features and deep semantic features to obtain an effective feature representation. The image generation unit is used to generate and stitch together super-resolution images based on the effective feature representation and the super-resolution reconstruction simulation dataset, generating high-resolution remote sensing super-resolution images that are spatially continuous, have complete bands, and meet the application standards of remote sensing images.

[0007] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for super-resolution reconstruction of mangrove optical remote sensing images.

[0008] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for super-resolution reconstruction of mangrove optical remote sensing images.

[0009] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for super-resolution reconstruction of mangrove optical remote sensing images.

[0010] According to the specific embodiments provided in this application, this application achieves the following technical effects: For the super-resolution reconstruction of mangrove optical remote sensing images, the reconstruction effect is significantly improved through dedicated data processing and deep learning network structure design. Specifically, a super-resolution reconstruction simulation dataset adapted to the characteristics of mangroves is first constructed to provide high-quality basic samples for model training. A modular super-resolution network combining feature extraction, feature fusion, and image generation is adopted, which can simultaneously and accurately extract shallow detail features and deep semantic features, avoiding information loss caused by single feature extraction. Effective feature expression is enhanced through adaptive weighted fusion, improving feature utilization. High-quality super-resolution image output is then achieved through image generation and stitching, and a stable and convergent dedicated model is obtained through iterative optimization. Overall, it can directly and efficiently super-reconstruct low-resolution mangrove remote sensing images to obtain high-resolution images, enhancing image texture and detail clarity, improving the accuracy of mangrove remote sensing information interpretation, providing more reliable data support for subsequent mangrove monitoring, analysis, and application. Furthermore, the model is highly targeted, generalizable, and practical. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is an application environment diagram of a mangrove optical remote sensing image super-resolution reconstruction method according to an embodiment of this application.

[0013] Figure 2 This is a flowchart illustrating a method for super-resolution reconstruction of mangrove optical remote sensing images, provided as an embodiment of this application.

[0014] Figure 3 This is a schematic diagram of the feature extraction module in one embodiment of this application.

[0015] Figure 4 This is a schematic diagram of the structure of the residual block in one embodiment of this application.

[0016] Figure 5 This is a schematic diagram of a switchable dilated convolution structure in one embodiment of this application.

[0017] Figure 6 This is a schematic diagram of the feature fusion module in one embodiment of this application.

[0018] Figure 7 This is a schematic diagram of the pixel attention mechanism in one embodiment of this application.

[0019] Figure 8 This is a schematic diagram of a mangrove optical remote sensing image super-resolution reconstruction device provided in an embodiment of this application.

[0020] Figure 9 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] To obtain more accurate information on mangrove distribution, this application proposes a lightweight super-resolution reconstruction model suitable for optical remote sensing images, which combines the pixel attention mechanism (PAM). This model addresses the issues of limited spatial resolution and inaccurate identification when identifying mangroves based on medium- and low-resolution optical remote sensing images.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] The super-resolution reconstruction method for mangrove optical remote sensing images provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be set up independently, integrated into server 102, or placed in the cloud or on another server. Terminal 101 can send high-resolution and low-resolution optical remote sensing images of the mangrove area at the average lowest tide time to server 102. Server 102 performs downscaling processing on the high-resolution and low-resolution optical remote sensing images respectively, and uses a pre-trained remote sensing image super-resolution reconstruction model to generate a high-resolution remote sensing super-resolution image of the mangrove area. Server 102 can then feed back the generated high-resolution remote sensing super-resolution image to terminal 101.

[0025] The terminal 101 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 102 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0026] In one exemplary embodiment, such as Figure 2 As shown, a method for super-resolution reconstruction of mangrove optical remote sensing images is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 102 as an example, the explanation includes the following steps 201 to 205.

[0027] Step 201: Acquire optical remote sensing images of mangroves and perform data preprocessing to construct a super-resolution reconstruction simulation dataset.

[0028] In a specific application example, step 201 includes steps 11 to 15.

[0029] Step 11: Collect high-resolution optical remote sensing images (HR) and low-resolution optical remote sensing images (LR) of the mangrove area at the average lowest tide time.

[0030] Step 12: Preprocess the acquired high-resolution optical remote sensing image (HR) and low-resolution optical remote sensing image (LR). The preprocessing includes radiometric calibration, atmospheric correction, geometric correction, image mosaicking, image registration, and image cropping to obtain high-resolution and low-resolution corrected images with fully aligned geographic coverage areas.

[0031] Step 13: Downscale the high-resolution corrected image and the low-resolution corrected image respectively. Use Gaussian filtering and downsampling to simulate the degradation mode of mangrove remote sensing image to obtain the corresponding high-resolution degradation sample and low-resolution degradation sample.

[0032] Step 14: Use a non-overlapping sliding window to perform batch cropping on the high-resolution degradation samples and low-resolution degradation samples to generate high-resolution degradation image blocks and low-resolution degradation image blocks. The cropping window size of the non-overlapping sliding window on the low-resolution degradation samples is h×w, and the cropping window size on the high-resolution degradation samples is (r·h)×(r·w), where h and w represent the height and width of the non-overlapping sliding window, respectively, and r represents the super-resolution magnification factor.

[0033] Step 15: Divide the high-resolution degraded image patch and the low-resolution degraded image patch into training set and validation set respectively according to the ratio. The high-resolution degraded image patch and the low-resolution degraded image patch correspond one-to-one to form a pair and together constitute the super-resolution reconstruction simulation dataset.

[0034] Step 202: Input the super-resolution reconstruction simulation dataset into the super-resolution network, which includes a feature extraction module, a feature fusion module, and an image generation module.

[0035] Step 203: Obtain the shallow features and deep semantic features of the super-resolution reconstruction simulation dataset through the feature extraction module.

[0036] In a specific application example, such as Figure 3 As shown, step 203 includes steps 31 to 34.

[0037] Step 31: Perform an interpolation upsampling operation on the low-resolution degraded image block to make its image size consistent with the size of the high-resolution degraded image block, thereby obtaining an upsampled image of the low-resolution degraded image block.

[0038] Step 32: The upsampled image of the low-resolution degraded image patch is concatenated with the high-resolution degraded image patch along the channel dimension. Shallow features of the super-resolution reconstruction simulation dataset are extracted through convolution operation. The convolution kernel of the convolution operation is 3×3.

[0039] Step 33: Based on the shallow features, extract mangrove semantic features of different depths through multiple sets of concatenated improved residual blocks.

[0040] like Figure 4 As shown, each improved residual block includes sequentially connected Switchable Atrous Convolution (SA_Conv), a leaky relu activation function, and a scaling layer. Each improved residual block also incorporates an external skip connection, which element-wise adds the input features to the output features of the residual branch to preserve low-level details and mitigate the vanishing gradient problem. In essence, each residual block internally performs deep processing of mangrove features through switchable atrous convolution, activation functions, and feature scaling, while utilizing residual connections to avoid gradient vanishing, gradually extracting more abstract semantic features and detailed information about mangroves.

[0041] like Figure 5As shown, branch 1 of the switchable dilated convolution is a 3×3 convolution with an atrous dilation of 1, used to capture local detailed features; branch 3 is a 3×3 convolution with an atrous dilation of 3, used to capture a wider range of contextual information; branch 2 consists of 5×5 average pooling and 1×1 convolution, used to generate the dynamic weights S and 1-S of branches 1 and 3.

[0042] Step 34: The mangrove semantic features of different depths output by multiple sets of improved residual blocks are spliced ​​and fused along the channel dimension. Then, the features are integrated and the dimensions are regularized by convolution operation with a kernel size of 3×3. Finally, deep semantic features that fuse multi-scale semantic information are obtained.

[0043] Step 204: The shallow features and deep semantic features are adaptively weighted and fused through the feature fusion module to obtain an effective feature representation.

[0044] In a specific application example, step 204 includes steps 41 to 42.

[0045] Step 41: Fuse the shallow features with the deep semantic features and input them into the self-constructed feature fusion module.

[0046] Step 42: The shallow features and deep semantic features are fused using the feature fusion module at the pixel level to obtain a discriminative and effective feature representation.

[0047] like Figure 6 As shown, the feature fusion module consists of a first convolutional layer (Conv), a first activation function layer (leaky_relu), a pixel attention mechanism layer (PAM), a second activation function layer (leaky_relu), a second convolutional layer (Conv), a third activation function layer (leaky_relu), a third convolutional layer (Conv), and a feature scaling layer (Scaling) connected in series.

[0048] like Figure 7 As shown, the pixel attention mechanism adopts a gated residual structure, generates pixel-wise attention weights through 1×1 convolution and Sigmoid activation, and multiplies them element-wise with the original input features to adaptively weight the spatial location features.

[0049] Step 205: The image generation module generates and stitches together super-resolution images based on the effective feature representation and the super-resolution reconstruction simulation dataset to obtain a high-resolution remote sensing super-resolution image that is spatially continuous, has complete bands, and meets the application standards of remote sensing images.

[0050] In a specific application example, step 205 includes steps 51 to 52.

[0051] Step 51: The effective feature representation is superimposed element-wise with the upsampled image of the low-resolution degraded image patch to achieve cross-stage feature mapping and information complementarity, thereby generating a high-resolution super-resolution image patch.

[0052] The image generation module is used to add the residual between the effective feature representation and the upsampled image of the low-resolution degraded image block, thereby realizing cross-stage feature mapping and information complementarity to obtain a super-resolution image block.

[0053] Step 52: Following the reverse order of the previous batch cropping, the high-resolution super-resolution image blocks are stitched and reassembled to restore the original band order and complete spatial dimensions, resulting in a spatially continuous, band-complete, and high-resolution remote sensing super-resolution image that meets the application standards of remote sensing images.

[0054] In this application, high-resolution remote sensing super-resolution imagery can achieve the same spatial resolution as the original high-resolution remote sensing imagery for the target low-resolution remote sensing imagery to be reconstructed. For example, if the spatial resolution of the original high-resolution imagery is 10m, the spatial resolution of the target low-resolution imagery to be reconstructed is 30m, and the spatial resolution of the high-resolution remote sensing super-resolution imagery is 10m.

[0055] This application introduces switchable dilated convolution in deep feature extraction, controlling the receptive field size by adjusting the dilation rate to adaptively capture mangrove patches at different scales. Furthermore, the parallel feature extraction and dilation design of the switchable dilated convolution better preserve the edge and texture information of the mangroves, avoiding detail loss. Moreover, the introduction of a pixel attention mechanism in feature fusion further weights the feature maps, strengthening key features, suppressing noise, and improving feature discriminativity. This allows the remote sensing image super-resolution reconstruction model to automatically focus on important regions, thereby improving feature extraction capabilities and super-resolution reconstruction accuracy.

[0056] This application addresses the problem of inaccurate mangrove patch boundary delineation and the easy omission of small patches and sparse mangroves in mangrove identification tasks due to insufficient spatial resolution of low- and medium-resolution optical remote sensing images. It proposes a super-resolution reconstruction method for mangrove optical remote sensing images based on a super-resolution reconstruction model. This method achieves super-resolution reconstruction by preprocessing, downscaling, batch cropping, data partitioning, feature extraction, feature fusion, image generation, and stitching / reassembly of low-resolution mangrove optical remote sensing images, ultimately outputting a super-resolution mangrove optical remote sensing image. This enables accurate identification and information extraction of mangrove features, improving the recognition accuracy of mangrove boundary information, small patches of mangroves, and sparse mangroves.

[0057] In summary, this application selects optical remote sensing images taken at the average lowest tide time to reduce tidal interference, ensuring the integrity and accuracy of mangrove land cover information and improving data reliability. High- and low-resolution dual remote sensing images are introduced as input, and degraded samples are constructed through preprocessing and downscaling. After batch cropping and data partitioning, a dedicated simulated dataset for mangrove super-resolution reconstruction training is built. A deep learning super-resolution model based on residual networks is employed. A hierarchical feature extraction strategy is used to first obtain shallow features, and then improved residual blocks composed of multiple concatenated modules are used to extract mangrove semantic features at different depths. The semantic features at different depths output from multiple residual blocks are concatenated and fused along the channel dimension. Finally, a 3×3 convolutional operation is used for feature integration and dimensionality regularization, resulting in deep semantic features of mangroves that integrate multi-scale semantic information. By combining pixel attention mechanisms and other operations, pixel-level adaptive weighted fusion of shallow and deep semantic features of mangroves is performed to enhance effective feature representation, suppress redundant information, and improve image texture and edge clarity. Finally, a high-resolution super-resolution image patch is generated through a residual addition structure to reduce information loss. Following the reverse order of the initial batch cropping, the super-resolution image patches are stitched and reassembled to generate a spatially continuous, band-complete, high-resolution remote sensing super-resolution image that meets remote sensing image application standards. The pre-trained network can be directly applied to acquire low-resolution remote sensing images of the target to be reconstructed, ultimately outputting a complete high-resolution remote sensing super-resolution image. This high-resolution remote sensing super-resolution image can provide clearer and more accurate remote sensing data support for mangrove resource monitoring and ecological protection.

[0058] Based on the same inventive concept, this application also provides a mangrove optical remote sensing image super-resolution reconstruction device for implementing the aforementioned mangrove optical remote sensing image super-resolution reconstruction method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more mangrove optical remote sensing image super-resolution reconstruction device embodiments provided below can be found in the limitations of the mangrove optical remote sensing image super-resolution reconstruction method described above, and will not be repeated here.

[0059] In one exemplary embodiment, such as Figure 8 As shown, a mangrove optical remote sensing image super-resolution reconstruction device is provided, comprising: a dataset construction unit 801, a feature extraction unit 802, a feature fusion unit 803, and an image generation unit 804; the dataset construction unit 801, the feature extraction unit 802, the feature fusion unit 803, and the image generation unit 804 are sequentially connected in communication to collaboratively complete the mangrove super-resolution image reconstruction task, with the specific functions as follows.

[0060] The dataset construction unit 801 is used to acquire optical remote sensing images of mangroves and perform data preprocessing to construct a standardized super-resolution reconstruction simulation dataset adapted to the input of the super-resolution network.

[0061] The feature extraction unit 802 is used to obtain the shallow features and deep semantic features of the super-resolution reconstruction simulation dataset.

[0062] The feature fusion unit 803 is used to adaptively weight and fuse the shallow features and deep semantic features to obtain an effective feature representation.

[0063] The image generation unit 804 is used to generate and stitch together super-resolution images based on the effective feature representation and the super-resolution reconstruction simulation dataset to obtain a high-resolution remote sensing super-resolution image that is spatially continuous, has complete bands, and meets the application standards of remote sensing images.

[0064] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 9 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores high-resolution and low-resolution optical remote sensing images of the mangrove area at the average lowest tide time. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for super-resolution reconstruction of mangrove optical remote sensing images.

[0065] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment to which the present application is applied. Specific computer equipment may include, for example, [the following is a list of possible additional structures]. Figure 9 The diagram shows more or fewer components, or combinations of certain components, or different component arrangements.

[0066] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0067] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0068] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0069] In this application, all actions to acquire signals, information, or data are carried out in compliance with the relevant data protection laws and policies of the country where the location is situated, and with the authorization granted by the owner of the relevant device.

[0070] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0071] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, etc., and are not limited to these.

[0072] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0073] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for super-resolution reconstruction of mangrove optical remote sensing images, characterized in that, include: High-resolution and low-resolution optical remote sensing images of the mangrove area at the average lowest tide time were collected. The high-resolution and low-resolution optical remote sensing images are preprocessed, including radiometric calibration, atmospheric correction, geometric correction, image mosaicking, image registration, and image cropping, to obtain high-resolution and low-resolution corrected images with fully aligned geographic coverage areas. The high-resolution corrected image and the low-resolution corrected image are downscaled respectively. Gaussian filtering and downsampling are used to simulate the degradation mode of mangrove remote sensing images to obtain corresponding high-resolution degradation samples and low-resolution degradation samples. The high-resolution degradation samples and low-resolution degradation samples are batch-cropped using a non-overlapping sliding window to generate high-resolution degradation image patches and low-resolution degradation image patches. The cropping window size of the non-overlapping sliding window on the low-resolution degradation sample is h×w, and the cropping window size on the high-resolution degradation sample is (r·h)×(r·w), where h and w represent the height and width of the non-overlapping sliding window, respectively, and r represents the super-resolution magnification factor. The high-resolution degraded image patches and low-resolution degraded image patches are divided into training set and validation set according to the ratio. The high-resolution degraded image patches and low-resolution degraded image patches correspond one-to-one to form a pair and together constitute the super-resolution reconstruction simulation dataset. The super-resolution reconstruction simulation dataset is input into the super-resolution network for model training and optimization to obtain the trained super-resolution reconstruction model. The trained super-resolution reconstruction model is used for super-resolution reconstruction of any mangrove forest image. The super-resolution network includes a feature extraction module, a feature fusion module, and an image generation module. The feature extraction module obtains shallow features and deep semantic features of the super-resolution reconstruction simulation dataset, specifically including: performing interpolation upsampling operation on the low-resolution degraded image patch to make its image size consistent with the size of the high-resolution degraded image patch, thereby obtaining an upsampled image of the low-resolution degraded image patch; The upsampled image of the low-resolution degraded image patch is concatenated with the high-resolution degraded image patch along the channel dimension. Shallow features of the super-resolution reconstruction simulation dataset are extracted through convolution operation, where the convolution kernel is 3×3. Based on the shallow features, semantic features of mangroves at different depths are extracted through multiple sets of concatenated improved residual blocks. Each improved residual block consists of a switchable dilated convolution, an activation function, and a feature scaling layer connected in sequence. Each improved residual block is equipped with an external skip connection, which adds the input features to the output features of the residual branch element by element to preserve the low-level details and alleviate the gradient vanishing problem. The semantic features of mangrove forests at different depths output by multiple sets of improved residual blocks are spliced ​​and fused along the channel dimension. Then, the features are integrated and the dimensions are regularized by convolution operation with a kernel size of 3×3. Finally, deep semantic features that fuse multi-scale semantic information are obtained. The feature fusion module adaptively weights and fuses the shallow features and deep semantic features to obtain an effective feature representation. Specifically, this includes fusing the shallow features and deep semantic features and inputting the fusion into the self-constructed feature fusion module. The feature fusion module consists of a first convolutional layer, a first activation function layer, a pixel attention mechanism layer, a second activation function layer, a second convolutional layer, a third activation function layer, a third convolutional layer, and a feature scaling layer connected in series. The feature fusion module performs pixel-level adaptive weighted fusion of the shallow features and deep semantic features to obtain a discriminative and effective feature representation. The image generation module generates and stitches together super-resolution images based on the effective feature representation and the super-resolution reconstruction simulation dataset to obtain a spatially continuous, band-complete, and high-resolution remote sensing super-resolution image that meets the application standards of remote sensing images. Specifically, this includes: element-wise superimposing the effective feature representation with the upsampled image of the low-resolution degraded image patch to generate a high-resolution super-resolution image patch; stitching together the high-resolution super-resolution image patch in the reverse order of the previous batch cropping to restore the original band order and complete spatial size, thus obtaining a spatially continuous, band-complete, and high-resolution remote sensing super-resolution image that meets the application standards of remote sensing images.

2. A device for super-resolution reconstruction of mangrove optical remote sensing images, performing the super-resolution reconstruction method for mangrove optical remote sensing images as described in claim 1, characterized in that, The mangrove optical remote sensing image super-resolution reconstruction device includes a dataset construction unit, a feature extraction unit, a feature fusion unit, and an image generation unit. These units are sequentially connected and collaboratively complete the mangrove super-resolution image reconstruction task. Their specific functions are as follows: The dataset construction unit is used to acquire optical remote sensing images of mangroves and perform data preprocessing to construct a standardized super-resolution reconstruction simulation dataset adapted to the input of the super-resolution network. The feature extraction unit is used to obtain the shallow features and deep semantic features of the super-resolution reconstruction simulation dataset; The feature fusion unit is used to adaptively weight and fuse the shallow features and deep semantic features to obtain an effective feature representation. The image generation unit is used to generate and stitch together super-resolution images based on the effective feature representation and the super-resolution reconstruction simulation dataset to obtain high-resolution remote sensing super-resolution images that are spatially continuous, have complete bands, and meet the application standards of remote sensing images.

3. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the mangrove optical remote sensing image super-resolution reconstruction method as described in claim 1.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the super-resolution reconstruction method for mangrove optical remote sensing images as described in claim 1.

5. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the super-resolution reconstruction method for mangrove optical remote sensing images as described in claim 1.

Images