A method and system for evaluating the effectiveness of mangrove afforestation based on deep learning

Abstract

The present invention discloses a method and system for evaluating the effectiveness of mangrove afforestation based on deep learning. The method includes: obtaining a digital orthophoto image and calculating the proportion of vegetation greenness area; using the YOLOv8 network model to identify mangrove targets in the mangrove restoration target area corresponding to less than the preset vegetation greenness area proportion, so as to obtain the number of mangrove trees per unit area; using the SegNet-SAM network model to extract the mangrove range in the mangrove restoration target area corresponding to greater than or equal to the preset vegetation greenness area proportion, so as to obtain the mangrove coverage; and evaluating the effectiveness of mangrove afforestation by combining the number of mangrove trees per unit area and the mangrove coverage results. The present invention can effectively improve the monitoring accuracy and efficiency of the effectiveness of mangrove afforestation. As a method and system for evaluating the effectiveness of mangrove afforestation based on deep learning, the present invention can be widely used in the field of forestry mapping remote sensing technology.

Description

Technical Field

[0001] The present invention relates to the technical field of forestry mapping and remote sensing, and particularly to a method and system for evaluating the effectiveness of mangrove afforestation based on deep learning. Background Art

[0002] At present, the evaluation of the effectiveness of mangrove afforestation mainly relies on the traditional on-site quadrat survey method, that is, the evaluation of the effectiveness of afforestation is carried out by manually measuring parameters such as the area, coverage, and the number of mangrove trees per unit area of mangroves on site. However, the traditional on-site quadrat survey method requires a large amount of manpower and material resources, and the problem of missing survey data will occur because some tidal flats are difficult for people or vehicles to reach. With the continuous progress of unmanned aerial vehicle (UAV) technology, UAVs have been widely used in the field of ecological surveys. However, ecological surveys based on UAV data still rely heavily on manual visual interpretation methods, and the accuracy of the results is limited by the experience of visual interpreters, and the data processing efficiency needs to be improved. In recent years, deep learning methods have been continuously developed and widely used in fields such as agriculture and industry. Mangrove afforestation and restoration projects are generally distributed in estuaries or intertidal zones with complex environments, affected by the ebb and flow of tides, and the morphological structures of mangroves in different afforestation areas vary greatly due to different mangrove species and planting times, which brings challenges to the remote sensing monitoring and evaluation of the effectiveness of mangrove afforestation. Summary of the Invention

[0003] In order to solve the above technical problems, the purpose of the present invention is to provide a method and system for evaluating the effectiveness of mangrove afforestation based on deep learning, which can select different deep learning methods through the vegetation greenness area ratio parameter, thereby improving the monitoring accuracy and evaluation efficiency of the effectiveness of mangrove afforestation.

[0004] The first technical solution adopted by the present invention is: A method for evaluating the effectiveness of mangrove afforestation based on deep learning, comprising the following steps:

[0005] Obtain the digital orthophoto image of the mangrove restoration target area and calculate the vegetation greenness area ratio to obtain the vegetation greenness area ratio data of the mangrove restoration target area;

[0006] Use the YOLOv8 network model to identify mangrove targets in the mangrove restoration target area corresponding to a vegetation greenness area ratio less than a preset value to obtain the number of mangrove trees per unit area;

[0007] Use the SegNet-SAM network model to extract the mangrove range in the mangrove restoration target area corresponding to a vegetation greenness area ratio greater than or equal to the preset value to obtain the mangrove coverage parameter;

[0008] Combine the number of mangrove trees per unit area and the mangrove coverage parameter to evaluate the effectiveness of mangrove afforestation and obtain the evaluation result of the effectiveness of mangrove afforestation.

[0009] Further, the step of obtaining the digital orthophoto image of the mangrove restoration target area and calculating the vegetation greenness area ratio to obtain the vegetation greenness area ratio data of the mangrove restoration target area specifically includes:

[0010] Obtain the visible light image of the mangrove restoration target area by drone;

[0011] Perform aerial triangulation and orthorectification processing on the visible light image of the mangrove restoration target area by drone in sequence to obtain the preliminary digital orthophoto image of the mangrove restoration target area;

[0012] Crop the preliminary digital orthophoto image of the mangrove restoration target area to obtain the digital orthophoto image of the mangrove restoration target area;

[0013] Based on the digital orthophoto image of the mangrove restoration target area, calculate the visible light band difference vegetation index to obtain the visible light band difference vegetation index data of the mangrove restoration target area;

[0014] Obtain the threshold of the visible light band difference vegetation index data of the mangrove restoration target area by the maximum inter-class variance method, and perform binarization processing on the visible light band difference vegetation index data to obtain the vegetation greenness map of the mangrove restoration target area;

[0015] Determine the vegetation greenness area of the mangrove restoration target area based on the vegetation greenness map of the mangrove restoration target area, and perform ratio calculation in combination with the area of the mangrove restoration target area to obtain the vegetation greenness area ratio data of the mangrove restoration target area.

[0016] Further, the step of performing mangrove target recognition on the mangrove restoration target area corresponding to the area ratio less than the preset vegetation greenness area ratio by the YOLOv8 network model to obtain the number of mangrove trees per unit area specifically includes:

[0017] Select the mangrove restoration target area corresponding to the area ratio less than the preset vegetation greenness area ratio as the first mangrove restoration target area;

[0018] Make a dataset from the digital orthophoto image of the first mangrove restoration target area and perform data preprocessing to obtain the first digital orthophoto image dataset;

[0019] Input the first digital orthophoto image dataset into the YOLOv8 network model, and the YOLOv8 network model includes a backbone network, a neck network, and a detection head network;

[0020] Based on the backbone network of the YOLOv8 network model, perform feature extraction processing on the first digital orthophoto image dataset to obtain the first mangrove image target feature;

[0021] The neck network based on the YOLOv8 network model performs multi-scale feature fusion on the target features of the first mangrove image to obtain the multi-scale fusion features of the first mangrove image target;

[0022] The detection head network based on the YOLOv8 network model performs target detection on the multi-scale fusion features of the first mangrove restoration image to obtain the target detection results of mangrove saplings;

[0023] Determine the number of mangrove saplings in the first mangrove restoration target area according to the target detection results of mangrove saplings, and perform proportional calculation in combination with the area of the first mangrove restoration target area to obtain the number of mangroves per unit area.

[0024] Further, the step of extracting the mangrove range of the mangrove restoration target area corresponding to the vegetation greenness area ratio greater than or equal to the preset value through the SegNet-SAM network model to obtain the mangrove cover parameter specifically includes:

[0025] Select the mangrove restoration target area corresponding to the vegetation greenness area ratio greater than or equal to the preset value as the second mangrove restoration target area;

[0026] Make a dataset of the digital orthophoto image of the second mangrove restoration target area and perform data preprocessing to obtain the second digital orthophoto image dataset;

[0027] Input the second digital orthophoto image dataset into the SegNet-SAM network model, and the SegNet-SAM network model includes a dual encoder module, a feature fusion module, and an improved decoder module;

[0028] Based on the dual encoder module of the SegNet-SAM network model, perform data encoding feature extraction processing on the second digital orthophoto image dataset to obtain a first feature map and a second feature map;

[0029] Based on the feature fusion module of the SegNet-SAM network model, perform feature fusion processing on the first feature map and the second feature map to obtain a fused feature map;

[0030] Based on the improved decoder module of the SegNet-SAM network model, perform decoding range extraction processing on the fused feature map and the second feature map to obtain the mangrove distribution result;

[0031] Determine the mangrove distribution range data based on the mangrove distribution result, and perform proportional calculation in combination with the area of the second mangrove restoration target area to obtain the mangrove cover parameter.

[0032] Further, the step of the dual-encoder module based on the SegNet-SAM network model performing data encoding feature extraction processing on the second digital orthophoto image dataset to obtain a first feature map and a second feature map specifically includes:

[0033] Input the second digital orthophoto image dataset into the dual-encoder module of the SegNet-SAM network model, and the dual-encoder module includes a SAM image encoder and a SegNet encoder;

[0034] Based on the SAM image encoder of the dual-encoder module, perform image encoding feature extraction processing on the second digital orthophoto image dataset to obtain a first feature map;

[0035] Based on the SegNet encoder of the dual-encoder module, perform image encoding feature extraction processing on the second digital orthophoto image dataset to obtain a second feature map.

[0036] Further, the step of the feature fusion module based on the SegNet-SAM network model performing feature fusion processing on the first feature map and the second feature map to obtain a fused feature map specifically includes:

[0037] Input the first feature map and the second feature map into the feature fusion module of the SegNet-SAM network model, and the feature fusion module includes a convolutional layer, a max pooling layer, a first channel splicing module, and a fusion convolutional module;

[0038] Based on the convolutional layer of the feature fusion module, perform channel dimension matching processing on the first feature map to obtain a preprocessed first feature map;

[0039] Based on the max pooling layer of the feature fusion module, perform downsampling processing on the second feature map to obtain a preprocessed second feature map;

[0040] Based on the first channel splicing module of the feature fusion module, perform channel splicing processing on the preprocessed first feature map and the preprocessed second feature map to obtain a spliced feature map;

[0041] Based on the fusion convolutional module of the feature fusion module, perform fusion convolutional processing on the spliced feature map to obtain a fused feature map.

[0042] Further, the step of the improved decoder module based on the SegNet-SAM network model performing decoding range extraction processing on the fused feature map and the second feature map to obtain the mangrove distribution result specifically includes:

[0043] Input the fused feature map and the second feature map into the improved decoder module of the SegNet-SAM network model. The improved decoder module includes a feature skip module, a transposed convolution upsampling module, a bilinear interpolation module, an adaptive skip connection module, a feature weighted fusion module, and a decoding convolution module;

[0044] Based on the feature skip module of the improved decoder module, perform multi-scale feature extraction processing on the second feature map to obtain a dimensional feature map, where the dimensional feature map includes a first dimensional feature map, a second dimensional feature map, and a third dimensional feature map;

[0045] Based on the transposed convolution upsampling module of the improved decoder module, perform transposed convolution upsampling processing on the fused feature map to obtain a preprocessed feature map;

[0046] Based on the bilinear interpolation module of the improved decoder module, perform interpolation processing on the dimensional feature map to obtain an interpolated dimensional feature map;

[0047] Based on the adaptive skip connection module of the improved decoder module, perform feature fusion on the interpolated dimensional feature map and the preprocessed feature map respectively to obtain a first fused feature map, a second fused feature map, and a third fused feature map;

[0048] Based on the feature weighted fusion module of the improved decoder module, perform weighted fusion on the first fused feature map, the second fused feature map, and the third fused feature map to obtain a weighted fused feature map;

[0049] Based on the decoding convolution module of the improved decoder module, perform decoding convolution processing on the weighted fused feature map to obtain the mangrove distribution result.

[0050] Furthermore, the step of performing multi-scale feature extraction processing on the second feature map by the feature skip module based on the improved decoder module to obtain a dimensional feature map specifically includes:

[0051] Input the second feature map into the feature skip module of the improved decoder module. The feature skip module includes a first branch, a second branch, and a third branch. The second branch includes a first convolutional layer and a first max pooling layer, and the third branch includes a second convolutional layer, a second max pooling layer, a third convolutional layer, and a third max pooling layer;

[0052] Based on the first branch of the feature skip module, obtain the second feature map and directly output it to get the first dimensional feature map;

[0053] Based on the second branch of the feature skip module, perform one-time convolution and pooling processing on the second feature map to obtain the second dimensional feature map;

[0054] Based on the third branch of the feature jump module, the second feature map is subjected to secondary convolution and pooling processing to obtain the third-dimensional feature map;

[0055] By combining the first-dimensional feature map, the second-dimensional feature map, and the third-dimensional feature map, a dimensional feature map is obtained.

[0056] Furthermore, the step of the adaptive jump connection module based on the improved decoder module for feature fusion of the interpolated dimensional feature map and the preprocessed feature map respectively to obtain the first fusion feature map, the second fusion feature map, and the third fusion feature map specifically includes:

[0057] Input the interpolated dimensional feature map and the preprocessed feature map into the adaptive jump connection module of the improved decoder module. The adaptive jump connection module includes a channel alignment module, a second channel splicing module, a hybrid attention mechanism module, a weight calculation module, and a weighted fusion module;

[0058] Based on the channel alignment module of the adaptive jump connection module, perform channel alignment processing on the preprocessed feature map to obtain the aligned feature map;

[0059] Based on the second channel splicing module of the adaptive jump connection module, perform channel splicing processing on the aligned feature map and the interpolated dimensional feature map to obtain the channel-spliced feature map;

[0060] Based on the hybrid attention mechanism module and the weight calculation module of the adaptive jump connection module, perform feature extraction and weight calculation on the channel-spliced feature map to obtain the weight value;

[0061] Based on the weighted fusion module of the adaptive jump connection module, perform weighted fusion on the aligned feature map and the interpolated dimensional feature map in combination with the weight value, and output the first fusion feature map, the second fusion feature map, and the third fusion feature map.

[0062] The second technical solution adopted by the present invention is: A mangrove afforestation effectiveness evaluation system based on deep learning, including:

[0063] The first module is used to obtain the digital orthophoto image of the mangrove restoration target area and calculate the vegetation greenness area ratio to obtain the vegetation greenness area ratio data of the mangrove restoration target area;

[0064] The second module is used to perform mangrove target recognition on the mangrove restoration target area corresponding to less than the preset vegetation greenness area ratio through the YOLOv8 network model to obtain the number of mangrove trees per unit area;

[0065] The third module is used to extract the mangrove range of the mangrove restoration target area corresponding to the vegetation greenness area ratio greater than or equal to the preset value through the SegNet-SAM network model, and obtain the mangrove coverage parameter;

[0066] The fourth module is used to evaluate the mangrove afforestation effect by combining the number of mangrove trees per unit area and the mangrove coverage parameter, and obtain the evaluation result of the mangrove afforestation effect.

[0067] The beneficial effects of the method and system of the present invention are as follows: By obtaining the digital orthophoto image of the mangrove restoration target area and calculating the vegetation greenness area ratio, the vegetation greenness area ratio data of the mangrove restoration target area is obtained. Then, the mangrove target is identified for the mangrove restoration target area corresponding to the vegetation greenness area ratio less than the preset value through the YOLOv8 network model, and the mangrove range is extracted for the mangrove restoration target area corresponding to the vegetation greenness area ratio greater than or equal to the preset value through the SegNet-SAM network model. Different deep learning methods are selected according to the vegetation greenness area ratio parameter to adapt to the monitoring and evaluation of mangrove afforestation and restoration projects under different mangrove growth states, ensuring the high accuracy of the evaluation result of the mangrove afforestation effect and the high efficiency of carrying out the evaluation work. Among them, the SegNet-SAM algorithm is used to extract the mangrove range in the mangrove restoration project area, which improves the robustness of the deep learning model in mangrove recognition in complex environments and the ability to retain fine boundary information of mangroves. Finally, the mangrove afforestation effect is evaluated by combining the number of mangrove trees per unit area and the mangrove coverage parameter, improving the monitoring accuracy and efficiency of the mangrove afforestation effect. Description of the Drawings

[0068] Figure 1 is the flowchart of the steps of a method for evaluating mangrove afforestation effect based on deep learning according to the present invention;

[0069] Figure 2 is the structural block diagram of a system for evaluating mangrove afforestation effect based on deep learning according to the present invention;

[0070] Figure 3 is the schematic diagram of the method steps for evaluating mangrove afforestation effect provided by a specific embodiment of the present invention;

[0071] Figure 4 is the schematic diagram of the method steps for obtaining the vegetation greenness area ratio provided by a specific embodiment of the present invention;

[0072] Figure 5 is the schematic diagram of the SegNet-SAM algorithm structure provided by a specific embodiment of the present invention;

[0073] Figure 6It is a schematic structural diagram of a fusion convolution module provided by a specific embodiment of the present invention;

[0074] Figure 7 It is a schematic structural diagram of a feature jump module provided by a specific embodiment of the present invention;

[0075] Figure 8 It is a schematic structural diagram of an adaptive jump connection module provided by a specific embodiment of the present invention;

[0076] Figure 9 It is a schematic structural diagram of a feature weighted fusion module provided by a specific embodiment of the present invention. Detailed implementation manners

[0077] The following further elaborates the present invention in detail in conjunction with the accompanying drawings and specific embodiments. For the step numbers in the following embodiments, they are only set for the convenience of explanation and illustration, and no limitation is imposed on the order between steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0078] Referring to Figure 1 , the present invention provides a method for evaluating the effectiveness of mangrove afforestation based on deep learning, and the method includes the following steps:

[0079] S100. Obtain the digital orthophoto image of the mangrove restoration target area and calculate the vegetation greenness area ratio to obtain the vegetation greenness area ratio data of the mangrove restoration target area;

[0080] S110. Obtain the visible light image of the mangrove restoration target area by using an unmanned aerial vehicle;

[0081] S120. Perform aerial triangulation and orthorectification processing on the visible light image of the mangrove restoration target area obtained by the unmanned aerial vehicle in sequence to obtain the preliminary digital orthophoto image of the mangrove restoration target area;

[0082] S130. Crop the preliminary digital orthophoto image of the mangrove restoration target area to obtain the digital orthophoto image of the mangrove restoration target area;

[0083] In this embodiment, as Figure 3As shown, a drone equipped with GPS RTK and the mapping aerial photography mode are used to collect visible light images at low tide in the area of the mangrove restoration project. The forward overlap is 60%-80%, the side overlap is 60%-80%, and the spatial resolution of the aerial drone images is better than 5 cm. Based on the drone images in the area of the mangrove restoration project, aerial triangulation and orthorectification processing are carried out to obtain digital orthophotos. The digital orthophotos of the mangrove restoration project area are cropped using the range data of the mangrove restoration project area to obtain the digital orthophotos of the mangrove restoration project area. Among them, aerial triangulation and orthorectification are processed through drone data processing software.

[0084] S140. Calculate the visible light band difference vegetation index based on the digital orthophotos of the mangrove restoration target area to obtain the visible light band difference vegetation index data of the mangrove restoration target area;

[0085] In this embodiment, as Figure 4 shown, based on the visible light digital orthophotos of the mangrove restoration project area, calculate the visible light band difference vegetation index VDVI (visible-band difference vegetation index), and the formula is as follows:

[0086] ;

[0087] Among them, is the pixel value of the green band, is the pixel value of the red band, is the pixel value of the blue band.

[0088] S150. Obtain the threshold of the visible light band difference vegetation index of the mangrove restoration target area by the maximum inter-class variance method, and perform binarization processing on the visible light band difference vegetation index data to obtain the vegetation greenness map of the mangrove restoration target area;

[0089] In this embodiment, the maximum inter-class variance method is used to process the visible light band difference vegetation index of the mangrove restoration project area to obtain the threshold , and use to perform binarization processing on the visible light band difference vegetation index data of the mangrove restoration project area to obtain the vegetation greenness map of the mangrove restoration project area, and calculate the vegetation greenness area .

[0090] S160. Determine the vegetation greenness area of the mangrove restoration target area based on the vegetation greenness map of the mangrove restoration target area and perform proportional calculation in combination with the area of the mangrove restoration target area to obtain the vegetation greenness area ratio data of the mangrove restoration target area.

[0091] In this embodiment, based on the data of the area range of the mangrove restoration project, the area of the area range of the mangrove restoration project is calculated. The proportion of the vegetation greenness area is calculated. .

[0092] S200. Use the YOLOv8 network model to identify mangrove targets in the mangrove restoration target area corresponding to a vegetation greenness area proportion less than the preset value, and obtain the number of mangrove trees per unit area.

[0093] First of all, it should be noted that considering that the evaluation of the effectiveness of mangrove afforestation is mainly based on the number of mangrove trees per unit area or the mangrove coverage, and either one can be used to judge whether the afforestation effectiveness is significant. However, due to the differences in the time of afforestation, tree species, etc. in the mangrove afforestation and restoration projects, the growth status of mangroves varies greatly. In some projects, the mangroves are already shaded by green trees, and the coverage parameter can be used for evaluation; while in some projects, the mangroves lack greenery, and only the number of mangrove trees per unit area can be used for evaluation. In this embodiment, the growth status of mangroves in the mangrove afforestation and restoration projects is judged through the vegetation greenness area proportion parameter, and the appropriate evaluation parameter (the number of mangrove trees per unit area or the mangrove coverage) is automatically selected to evaluate its afforestation effectiveness, adapting to the monitoring and evaluation of mangrove afforestation and restoration projects under different mangrove growth states, ensuring the high precision of the evaluation results of mangrove afforestation effectiveness and the high efficiency of carrying out the evaluation work.

[0094] Different deep learning methods are selected using the vegetation greenness area proportion: when the vegetation greenness area proportion is less than 20%, it indicates that the mangroves in the mangrove restoration project area are mainly saplings and the proportion of green leaves is not high. The YOLOv8 algorithm is selected for mangrove target recognition, and the number of mangrove trees per unit area is calculated; when the vegetation greenness area proportion is not less than 20%, it indicates that the proportion of green leaves of the mangroves in the mangrove restoration project area is relatively high, and even some areas may form forests. The mangrove coverage parameter is required to describe the growth status of mangroves. The SegNet-SAM algorithm is used for mangrove range extraction, and the mangrove coverage is calculated.

[0095] Specifically, select the mangrove restoration target area corresponding to the proportion of the preset vegetation greenness area as the first mangrove restoration target area; make a dataset from the digital orthophoto image of the first mangrove restoration target area and perform data preprocessing to obtain the first digital orthophoto image dataset; input the first digital orthophoto image dataset into the YOLOv8 network model, where the YOLOv8 network model includes a backbone network, a neck network, and a detection head network; based on the backbone network of the YOLOv8 network model, perform feature extraction processing on the first digital orthophoto image dataset to obtain the first mangrove image target features; based on the neck network of the YOLOv8 network model, perform multi-scale feature fusion on the first mangrove image target features to obtain the first mangrove image target multi-scale fusion features; based on the detection head network of the YOLOv8 network model, perform target detection on the multi-scale fusion features of the first mangrove restoration image to obtain the target detection results of mangrove saplings; determine the number of mangrove saplings in the first mangrove restoration target area according to the target detection results of mangrove saplings, and perform ratio calculation in combination with the area of the first mangrove restoration target area to obtain the number of mangrove trees per unit area.

[0096] In this embodiment, the visible light digital orthophoto image of the mangrove restoration project area is cropped into sub-image blocks of the same size (the side length of the sub-image block is defaulted to 640 pixels), select the sub-image blocks containing at least one complete mangrove sapling, and randomly set the sub-image blocks as the training set, validation set, and test set according to the ratio of 7:2:1. For the training set and validation set, use an image marking tool to mark the mangrove saplings in the sub-image blocks, and store the information such as the position and border size of the mangrove saplings in a txt file in YOLO format. For the sub-image blocks in the training set, expand the dataset through five common operations: rotation, flipping, brightness adjustment, contrast adjustment, and mosaic.

[0097] Select the YOLOv8 model and input the training set into the model for training. Input the validation set into the neural network model saved after training for detection, and evaluate the detection accuracy of different neural network models, and select the model with the highest average precision as the detection model.

[0098] Select Precision and Recall to evaluate the performance of the model. Precision reflects the proportion of correctly predicted positive samples in all predicted positive samples by the model, and Recall is the proportion of correctly predicted positive samples in all actual positive samples by the model. The specific formulas are as follows:

[0099] ;

[0100] Among them, represents the number of true mangrove saplings (true positives) detected; The number of samples that represent misidentifying non-mangrove saplings as mangrove saplings (false positives); The number of samples that represent misdetecting true mangrove saplings as non-mangrove saplings (false negatives).

[0101] Use the detection model to detect mangrove saplings in the test set, obtain the bounding box and confidence detection data of mangrove saplings, summarize the mangrove sapling detection data of the sub-image blocks of the visible light digital orthophoto image in the mangrove restoration project area, and obtain the number of mangrove saplings in the mangrove restoration project area , combined with the area of the mangrove restoration project area , calculate the number of mangrove trees per unit area .

[0102] S300. Extract the mangrove range of the mangrove restoration target area corresponding to a vegetation greenness area ratio greater than or equal to the preset value through the SegNet-SAM network model to obtain the mangrove cover parameter;

[0103] First of all, it should be noted that the network structure of the SegNet model includes an encoder and a decoder. The SegmentAnything Model (SAM) model includes an image encoder, a prompt encoder, and a mask decoder. The SAM model can segment specific targets from images through user prompts or can perform zero-shot segmentation of everything. The advantage of the SAM model is its extremely strong zero-shot generalization ability and its ability to accurately identify and segment objects in complex scenes. SAM itself is only a segmentation model and cannot obtain semantic information; while the SegNet model can achieve semantic segmentation, but the segmentation structure is not fine enough. This embodiment combines the advantages of the SAM and SegNet models respectively and proposes the SegNet-SAM algorithm, as Figure 5 shown, to extract the mangrove range of the mangrove restoration project area.

[0104] S310. Select the mangrove restoration target area corresponding to a vegetation greenness area ratio greater than or equal to the preset value as the second mangrove restoration target area;

[0105] S320. Make a dataset from the digital orthophoto image of the second mangrove restoration target area and perform data preprocessing to obtain the second digital orthophoto image dataset;

[0106] In this embodiment, the visible light digital orthophoto image of the mangrove restoration project area is cropped into sub-image blocks of the same size (the side length of the sub-image block is defaulted to 256 pixels), the sub-image blocks containing at least one complete mangrove sapling are selected, and the sub-image blocks are randomly set as the training set, validation set and test set according to the ratio of 7:2:1. For the training set and the validation set, the mangroves in the sub-image blocks are marked using an image marking tool. For the sub-image blocks in the training set, the data set is amplified through five common operations: rotation, flipping, brightness adjustment, contrast adjustment, and mosaic.

[0107] S330. Input the second digital orthophoto image data set into the SegNet-SAM network model, where the SegNet-SAM network model includes a dual encoder module, a feature fusion module, and an improved decoder module;

[0108] S340. Based on the dual encoder module of the SegNet-SAM network model, perform data encoding feature extraction processing on the second digital orthophoto image data set to obtain a first feature map and a second feature map;

[0109] Specifically, input the second digital orthophoto image data set into the dual encoder module of the SegNet-SAM network model. The dual encoder module includes a SAM image encoder and a SegNet encoder; based on the SAM image encoder of the dual encoder module, perform image encoding feature extraction processing on the second digital orthophoto image data set to obtain a first feature map; based on the SegNet encoder of the dual encoder module, perform image encoding feature extraction processing on the second digital orthophoto image data set to obtain a second feature map.

[0110] In this embodiment, the SegNet-SAM network is mainly composed of a dual encoder module, a feature fusion module, and an improved decoder module. The dual encoder module is composed of the encoder of the SAM model and the encoder of the SegNet model in parallel. The encoder of the SAM model processes the input data set and outputs a feature map ; the encoder of the SegNet model processes the input data set and outputs a feature map .

[0111] S350. Based on the feature fusion module of the SegNet-SAM network model, perform feature fusion processing on the first feature map and the second feature map to obtain a fused feature map;

[0112] Specifically, the first feature map and the second feature map are input into the feature fusion module of the SegNet-SAM network model. The feature fusion module includes a convolutional layer, a max pooling layer, a first channel concatenation module, and a fusion convolutional module. Based on the convolutional layer of the feature fusion module, channel dimension matching processing is performed on the first feature map to obtain a preprocessed first feature map. Based on the max pooling layer of the feature fusion module, downsampling processing is performed on the second feature map to obtain a preprocessed second feature map. Based on the first channel concatenation module of the feature fusion module, channel concatenation processing is performed on the preprocessed first feature map and the preprocessed second feature map to obtain a concatenated feature map. Based on the fusion convolutional module of the feature fusion module, fusion convolutional processing is performed on the concatenated feature map to obtain a fused feature map.

[0113] In this embodiment, the feature fusion module realizes the feature fusion of the feature maps output by the encoders of the SAM and SegNet models through convolution, max pooling, channel concatenation, and fusion convolution. The feature fusion module first performs downsampling on the feature map through max pooling, so that the size of the processed feature map is the same as the size of the feature map output by the SAM encoder ; at the same time, convolution is used to perform channel dimension matching on the feature map so that the number of channels of the processed feature map is the same as the number of channels of the feature map . The channel concatenation module is used to concatenate the feature map and the feature map in the channel dimension to fuse the feature data output by the two encoder branches of SAM and SegNet in the dual encoder module, and obtain the feature map ; then, the fusion convolutional module is used to process the feature map , and the feature map is output. Among them, the structure of the fusion convolutional module is as shown, and it is connected in sequence to depth convolution and Figure 6 point convolution. depth convolution and point convolution.

[0114] S360. The improved decoder module based on the SegNet-SAM network model performs decoding range extraction processing on the fused feature map and the second feature map to obtain the mangrove distribution result;

[0115] Specifically, the fused feature map and the second feature map are input into the improved decoder module of the SegNet-SAM network model. The improved decoder module includes a feature skip module, a transposed convolution upsampling module, a bilinear interpolation module, an adaptive skip connection module, a feature weighted fusion module, and a decoding convolution module. Based on the feature skip module of the improved decoder module, multi-scale feature extraction is performed on the second feature map to obtain a dimensional feature map, which includes a first dimensional feature map, a second dimensional feature map, and a third dimensional feature map. Based on the transposed convolution upsampling module of the improved decoder module, transposed convolution upsampling is performed on the fused feature map to obtain a preprocessed feature map. Based on the bilinear interpolation module of the improved decoder module, interpolation is performed on the dimensional feature map to obtain an interpolated dimensional feature map. Based on the adaptive skip connection module of the improved decoder module, feature fusion is respectively performed on the interpolated dimensional feature map and the preprocessed feature map to obtain a first fused feature map, a second fused feature map, and a third fused feature map. Based on the feature weighted fusion module of the improved decoder module, weighted fusion is performed on the first fused feature map, the second fused feature map, and the third fused feature map to obtain a weighted fused feature map. Based on the decoding convolution module of the improved decoder module, decoding convolution is performed on the weighted fused feature map to obtain the mangrove distribution result.

[0116] In this embodiment, the improved decoder module outputs the mangrove distribution map result through transposed convolution upsampling, a feature skip module, a bilinear interpolation module, an adaptive skip connection module, feature weighted fusion, and decoding convolution. The improved decoder module first performs transposed convolution upsampling on the feature map and outputs a feature map (with a dimension of ). The feature skip module is used to process the feature map , build a multi-scale feature bridge between the encoder and the decoder, and output three feature maps with different dimensions (with a dimension of 64×12×128 ), feature map (with a dimension of ), and feature map (with a dimension of ).

[0117] The bilinear interpolation is used to process the feature map , feature map , and feature map to align them with the feature map in space, and output three corresponding processed feature maps , feature map , and feature map . The feature map and the feature map are input into the adaptive skip connection module for fusing the features of the two models and outputting a feature map ; the feature map and the feature map are input into the adaptive skip connection module and a feature map is output ; the feature map and the feature map are input into the adaptive skip connection module and a feature map is output .

[0118] Finally, the decoding convolutional module is used to process the feature map , and the mangrove distribution range data of the mangrove restoration project area is output

[0119] Among them, for the feature skip module, it further includes inputting the second feature map into the feature skip module of the improved decoder module. The feature skip module includes a first branch, a second branch and a third branch. The second branch includes a first convolutional layer and a first max pooling layer. The third branch includes a second convolutional layer, a second max pooling layer, a third convolutional layer and a third max pooling layer; based on the first branch of the feature skip module, the second feature map is obtained and directly output to obtain a first-dimensional feature map; based on the second branch of the feature skip module, the second feature map is subjected to one-time convolution and pooling processing to obtain a second-dimensional feature map; based on the third branch of the feature skip module, the second feature map is subjected to two-time convolution and pooling processing to obtain a third-dimensional feature map; the first-dimensional feature map, the second-dimensional feature map and the third-dimensional feature map are combined to obtain a dimensional feature map

[0120] In this embodiment, the structure of the feature skip module is as Figure 7 shown, including 3 branches. The first branch directly outputs the feature map ; the second branch is sequentially connected to convolution and max pooling and outputs the feature map ; the third branch is sequentially connected to convolution, max pooling, convolution and max pooling and outputs the feature map .

[0121] Further, for the adaptive skip connection module, it further includes inputting the interpolated dimensional feature map and the preprocessed feature map into the adaptive skip connection module of the improved decoder module. The adaptive skip connection module includes a channel alignment module, a second channel splicing module, a hybrid attention mechanism module, a weight calculation module, and a weighted fusion module. Based on the channel alignment module of the adaptive skip connection module, channel alignment processing is performed on the preprocessed feature map to obtain an aligned feature map. Based on the second channel splicing module of the adaptive skip connection module, channel splicing processing is performed on the aligned feature map and the interpolated dimensional feature map to obtain a channel-spliced feature map. Based on the hybrid attention mechanism module and the weight calculation module of the adaptive skip connection module, feature extraction and weight calculation are performed on the channel-spliced feature map to obtain a weight value. Based on the weighted fusion module of the adaptive skip connection module, weighted fusion is performed on the aligned feature map and the interpolated dimensional feature map in combination with the weight value, and the first fusion feature map, the second fusion feature map, and the third fusion feature map are output.

[0122] The adaptive skip connection module is as Figure 8 shown. First, perform convolution on the feature map to align its number of channels with the number of channels of the feature map ( ), and output the feature map ; then splice the feature map and the feature map channel-wise, and output the feature map ; input the feature map into the hybrid attention mechanism and calculate the weights, that is, perform convolution, GeLU activation function processing, convolution, and Sigmoid activation function processing on the feature map in sequence to obtain the corresponding weight value ; then fuse the feature map and the feature map by weighted fusion, and output the feature map ( ).

[0123] The weight calculation formula in the adaptive skip connection module is as follows:

[0124] ;

[0125] where, is the input feature map; is convolution; is Convolution; GeLU is the GeLU (Gaussian Error Linear Unit) activation function; is the Sigmoid activation function.

[0126] The weighted fusion formula in the adaptive skip connection module is as follows:

[0127] ;

[0128] where is the feature map after the fusion of SAM and SegNet; is the SegNet feature map. In this embodiment, is ; is .

[0129] Use the feature weighted fusion module to perform weighted fusion on the feature map , the feature map and the feature map . The feature weighted fusion module is as shown in Figure 9 . By inputting the feature map , the feature map and the feature map through the input layer, first use the third-channel splicing module to splice the feature map , the feature map and the feature map in the channel dimension, and output the feature map ; then compress the spatial dimension to 1×1 through the AdaptiveAvgPool2d layer and generate a channel description vector. Subsequently, through the Flatten layer and the Linear layer, input the flattened vector into the dimensionality reduction linear layer for feature compression, and then enhance the non-linear expression ability through the ReLU activation function. Then, restore the original channel dimension through the dimensionality increase Linear layer, and finally generate a normalized channel weight mask through the Sigmoid activation function. The weight mask is used to scale the features of the original input feature map through per-channel multiplication to generate the calibrated output feature map .

[0130] S370. Determine the mangrove distribution range data based on the mangrove distribution results, and perform proportional calculation in combination with the area of the second mangrove restoration target area to obtain the mangrove cover parameter.

[0131] In this embodiment, based on the SegNet-SAM model, the training set is input into the model for training. The validation set is input into the trained neural network model for detection, and the detection accuracy of different neural network models is evaluated. The model with the highest average accuracy is selected as the detection model.

[0132] The performance of the model is evaluated using Precision and Recall. Precision reflects the proportion of correctly predicted positive samples among all samples predicted as positive, and Recall is the proportion of correctly predicted positive samples among all actual positive samples. The specific formulas are as follows:

[0133] ;

[0134] where represents the number of true mangrove (true positive) samples detected; represents the number of samples that wrongly identify non-mangroves as mangroves (false positives); represents the number of true mangroves misdetected as non-mangroves (false negatives).

[0135] The detection model is used to detect the mangrove distribution in the test set to obtain mangrove distribution range data. The mangrove distribution range detection data of the sub-image blocks of the visible light digital orthophoto image in the mangrove restoration project area is summarized to obtain the mangrove distribution range area of the mangrove restoration project area , combined with the area of the mangrove restoration project area , to calculate the mangrove coverage .

[0136] S400. Combine the number of mangrove trees per unit area and the mangrove coverage parameter to evaluate the effectiveness of mangrove afforestation and obtain the evaluation result of the effectiveness of mangrove afforestation.

[0137] Specifically, based on the calculated number of mangrove trees per unit area or the mangrove coverage parameter, the evaluation of the effectiveness of mangrove afforestation in the mangrove restoration project is carried out. When the number of mangrove trees per unit area trees / or the mangrove coverage ≥ 50%, the effectiveness of mangrove afforestation is significant; otherwise, the effectiveness of mangrove afforestation is not significant.

[0138] In summary, as Figure 3 shown, in the embodiments of the present invention, a visible light digital camera carried by a drone is used to obtain a high-resolution orthophoto image at low tide in the mangrove afforestation and restoration project area. First, different deep learning methods are used for the proportion of vegetation greenness area. When the proportion of vegetation greenness area is less than 20%, the YOLOv8 algorithm is selected for mangrove target recognition and the number of mangrove trees per unit area is calculated. When the proportion of vegetation greenness area is not less than 20%, the SegNet-SAM algorithm is used for mangrove range extraction and the mangrove coverage is calculated; finally, the effectiveness of mangrove afforestation is evaluated through the extracted mangrove parameters.

[0139] Therefore, the embodiments of the present invention have the following advantages compared with the prior art:

[0140] 1) Select different deep learning methods through the vegetation greenness area ratio parameter to adapt to the monitoring and evaluation of mangrove afforestation and restoration projects under different mangrove growth states, ensuring the high precision of the evaluation results of mangrove afforestation effectiveness and the high efficiency of carrying out the evaluation work.

[0141] 2) Combine the SAM and SegNet models to propose the SegNet-SAM algorithm for extracting the mangrove range in the mangrove restoration project area, improving the robustness of the deep learning model in mangrove recognition under complex environments and the ability to retain fine boundary information of mangroves.

[0142] 3) Use the combination of unmanned aerial vehicle aerial photography and deep learning methods to evaluate the effectiveness of mangrove afforestation, which can assist ecological protection and restoration experts in improving the efficiency of monitoring important mangrove parameters, reducing labor and time costs, and providing important data support for the management and planning of subsequent mangrove afforestation projects and the decision-making of mangrove protection and restoration.

[0143] Refer to Figure 2 , a mangrove afforestation effectiveness evaluation system based on deep learning, including:

[0144] The first module 201 is used to obtain the digital orthophoto of the mangrove restoration target area and calculate the vegetation greenness area ratio to obtain the vegetation greenness area ratio data of the mangrove restoration target area;

[0145] The second module 202 is used to identify mangrove targets in the mangrove restoration target area corresponding to a vegetation greenness area ratio less than the preset value through the YOLOv8 network model to obtain the number of mangrove trees per unit area;

[0146] The third module 203 is used to extract the mangrove range in the mangrove restoration target area corresponding to a vegetation greenness area ratio greater than or equal to the preset value through the SegNet-SAM network model to obtain the mangrove coverage parameter;

[0147] The fourth module 204 is used to combine the number of mangrove trees per unit area and the mangrove coverage parameter to evaluate the effectiveness of mangrove afforestation and obtain the evaluation result of mangrove afforestation effectiveness.

[0148] The content in the above method embodiments is applicable to the system embodiments. The functions specifically implemented by the system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0149] The above is a specific description of the preferred embodiment of the present invention. However, the present invention is not limited to the described embodiment. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A mangrove afforestation effectiveness evaluation method based on deep learning, characterized in that: The following steps are involved: Obtain digital orthophotos of the target area for mangrove restoration and calculate the vegetation green area ratio to obtain vegetation green area ratio data for the target area for mangrove restoration; The YOLOv8 network model is used to identify mangrove targets in the first mangrove restoration target area corresponding to the preset vegetation green area ratio, and the number of mangrove trees per unit area is obtained; The SegNet-SAM network model is used to extract the mangrove range of the second mangrove restoration target area corresponding to the preset vegetation green area ratio, and obtain the mangrove coverage parameter; Combining the number of mangrove trees per unit area with the mangrove coverage parameters, the effectiveness of mangrove afforestation was evaluated to obtain the results of mangrove afforestation effectiveness evaluation. The SegNet-SAM network model includes a dual encoder module, a feature fusion module and an improved decoder module, wherein the dual encoder module includes a SAM image encoder and a SegNet encoder, the feature fusion module includes a convolution layer, a maximum pooling layer, a first channel splicing module and a fusion convolution module, and the improved decoder module includes a feature skip module, a deconvolution upsampling module, a bilinear interpolation module, an adaptive skip connection module, a feature weighted fusion module and a decoding convolution module; The specific process of extracting the mangrove range in the second mangrove restoration target area is as follows: The digital orthophotos of the second mangrove restoration target area are used to prepare a data set and perform data preprocessing to obtain a second digital orthophoto data set; In the dual encoder module: performing image coding feature extraction processing on the second digital orthophoto dataset based on the SAM image encoder and the SegNet encoder respectively to obtain a first feature map and a second feature map; In the feature fusion module: the first feature map is processed based on the convolution layer, the second feature map is processed based on the maximum pooling layer, the first channel splicing module performs channel splicing processing on the processed first feature map and the second feature map, and then the fusion convolution module performs fusion convolution processing on the spliced ​​feature map to obtain a fused feature map; In the improved decoder module: Based on the deconvolution upsampling module, the fused feature map is deconvoluted and upsampled to obtain a preprocessed feature map; Performing multi-scale feature extraction processing on the second feature map based on the feature jump module to obtain a dimensional feature map, wherein the dimensional feature map includes a first dimensional feature map, a second dimensional feature map, and a third dimensional feature map; performing interpolation processing on the dimensional feature map based on the bilinear interpolation module to obtain an interpolated dimensional feature map; Based on the adaptive skip connection module, the interpolated dimensional feature map and the preprocessed feature map are fused to obtain three fused feature maps. Based on the feature weighted fusion module, the three fusion feature maps are weightedly fused to obtain a weighted fusion feature map; Based on the decoding convolution module, the weighted fused feature map is decoded and convolved to obtain the mangrove distribution results.

2. According to claim 1, a method for evaluating the effectiveness of mangrove afforestation based on deep learning is characterized in that: The step of obtaining a digital orthophoto of the target area for mangrove restoration and calculating the vegetation green area ratio to obtain vegetation green area ratio data for the target area for mangrove restoration specifically includes: Obtain drone visible light images of mangrove restoration target areas through drones; The visible light images of the mangrove restoration target area were processed by aerial triangulation and orthorectification in sequence to obtain a preliminary digital orthophoto of the mangrove restoration target area. The digital orthophoto of the preliminary mangrove restoration target area is cropped to obtain the digital orthophoto of the mangrove restoration target area; Based on the digital orthophoto of the mangrove restoration target area, the visible light band difference vegetation index is calculated to obtain the visible light band difference vegetation index data of the mangrove restoration target area; The threshold of the visible light band difference vegetation index data of the mangrove restoration target area was obtained by the maximum inter-class variance method, and the visible light band difference vegetation index data was binarized to obtain the vegetation greenness map of the mangrove restoration target area. Based on the vegetation greenness map of the mangrove restoration target area, the vegetation greenness area of ​​the mangrove restoration target area is determined, and the proportion is calculated based on the area of ​​the mangrove restoration target area to obtain the vegetation greenness area proportion data of the mangrove restoration target area.

3. According to claim 2, a method for evaluating the effectiveness of mangrove afforestation based on deep learning is characterized in that: The step of identifying mangrove targets in the first mangrove restoration target area corresponding to the preset vegetation green area ratio by using the YOLOv8 network model to obtain the number of mangrove trees per unit area specifically includes: Select a mangrove restoration target area that is smaller than a preset vegetation green area ratio as the first mangrove restoration target area; The digital orthophotos of the first mangrove restoration target area are used to prepare a data set and perform data preprocessing to obtain a first digital orthophoto data set; Inputting the first digital orthophoto dataset into a YOLOv8 network model, wherein the YOLOv8 network model includes a backbone network, a neck network, and a detection head network; Based on the backbone network of the YOLOv8 network model, feature extraction processing is performed on the first digital orthophoto dataset to obtain the first mangrove image target features; Based on the neck network of the YOLOv8 network model, multi-scale feature fusion is performed on the target features of the first mangrove image to obtain the multi-scale fusion features of the first mangrove image target; The detection head network based on the YOLOv8 network model performs target detection on the multi-scale fusion features of the first mangrove image target and obtains the target detection result of the mangrove saplings; The number of mangrove saplings in the first mangrove restoration target area is determined based on the target detection results of mangrove saplings, and the proportion is calculated based on the area of ​​the first mangrove restoration target area to obtain the number of mangroves per unit area.

4. According to claim 3, a method for evaluating the effectiveness of mangrove afforestation based on deep learning is characterized in that: The step of extracting the mangrove range of the second mangrove restoration target area corresponding to a vegetation green area ratio greater than or equal to a preset ratio by using the SegNet-SAM network model to obtain the mangrove coverage parameter specifically includes: Selecting a mangrove restoration target area corresponding to a vegetation green area ratio greater than or equal to a preset ratio as the second mangrove restoration target area; The digital orthophotos of the second mangrove restoration target area are used to prepare a data set and perform data preprocessing to obtain a second digital orthophoto data set; The second digital orthophoto dataset was input into the SegNet-SAM network model to obtain the mangrove distribution results; Based on the mangrove distribution results, the mangrove distribution range data was determined, and the proportion calculation was performed based on the area of ​​the second mangrove restoration target area to obtain the mangrove coverage parameters.

5. A method for evaluating the effectiveness of mangrove afforestation based on deep learning according to claim 4, characterized in that: The step of performing multi-scale feature extraction processing on the second feature map based on the feature skip module to obtain a dimensional feature map specifically includes: Inputting the second feature map to a feature skip module of the improved decoder module, the feature skip module comprising a first branch, a second branch and a third branch, the second branch comprising a first convolutional layer and a first maximum pooling layer, the third branch comprising a second convolutional layer, a second maximum pooling layer, a third convolutional layer and a third maximum pooling layer; Based on the first branch of the feature skip module, the second feature map is obtained and directly output to obtain a first dimension feature map; Based on the second branch of the feature skip module, a convolution and pooling process is performed on the second feature map to obtain a second dimensional feature map; Based on the third branch of the feature skip module, the second feature map is subjected to secondary convolution and pooling processing to obtain a feature map of the third dimension; The first dimension feature map, the second dimension feature map and the third dimension feature map are combined to obtain a dimensional feature map.

6. A method for evaluating the effectiveness of mangrove afforestation based on deep learning according to claim 5, characterized in that: The step of performing feature fusion on the interpolated dimensional feature map and the preprocessed feature map based on the adaptive skip connection module to obtain three fused feature maps specifically includes: Inputting the interpolated dimensional feature map and the preprocessed feature map into an adaptive skip connection module of the improved decoder module, wherein the adaptive skip connection module includes a channel alignment module, a second channel splicing module, a hybrid attention mechanism module, a weight calculation module and a weighted fusion module; A channel alignment module based on an adaptive skip connection module performs channel alignment on the preprocessed feature map to obtain an aligned feature map; A second channel splicing module based on an adaptive skip connection module performs channel splicing on the aligned feature map and the interpolated dimensional feature map to obtain a channel-spliced ​​feature map; Based on the hybrid attention mechanism module and weight calculation module of the adaptive skip connection module, feature extraction and weight calculation are performed on the feature map after channel splicing to obtain the weight value; Based on the weighted fusion module of the adaptive skip connection module, the aligned feature map and the interpolated dimensional feature map are weightedly fused in combination with the weight value, and the first fused feature map, the second fused feature map and the third fused feature map are output.

7. A mangrove afforestation effectiveness evaluation system based on deep learning, characterized in that: Includes the following modules: The first module is used to obtain the digital orthophoto of the mangrove restoration target area and calculate the vegetation green area ratio to obtain the vegetation green area ratio data of the mangrove restoration target area; The second module is used to identify mangrove targets in the first mangrove restoration target area corresponding to a preset vegetation green area ratio by using the YOLOv8 network model, and obtain the number of mangrove trees per unit area; The third module is used to extract the mangrove range of the second mangrove restoration target area corresponding to a preset vegetation green area ratio by using the SegNet-SAM network model to obtain the mangrove coverage parameter; The fourth module is used to evaluate the effectiveness of mangrove afforestation by combining the number of mangrove trees per unit area with the mangrove coverage parameter, and obtain the evaluation results of the effectiveness of mangrove afforestation; The SegNet-SAM network model includes a dual encoder module, a feature fusion module and an improved decoder module, wherein the dual encoder module includes a SAM image encoder and a SegNet encoder, the feature fusion module includes a convolution layer, a maximum pooling layer, a first channel splicing module and a fusion convolution module, and the improved decoder module includes a feature skip module, a deconvolution upsampling module, a bilinear interpolation module, an adaptive skip connection module, a feature weighted fusion module and a decoding convolution module; The specific process of extracting the mangrove range in the second mangrove restoration target area is as follows: The digital orthophotos of the second mangrove restoration target area are used to prepare a data set and perform data preprocessing to obtain a second digital orthophoto data set; In the dual encoder module: performing image coding feature extraction processing on the second digital orthophoto dataset based on the SAM image encoder and the SegNet encoder respectively to obtain a first feature map and a second feature map; In the feature fusion module: the first feature map is processed based on the convolution layer, the second feature map is processed based on the maximum pooling layer, the first channel splicing module performs channel splicing processing on the processed first feature map and the second feature map, and then the fusion convolution module performs fusion convolution processing on the spliced ​​feature map to obtain a fused feature map; In the improved decoder module: Based on the deconvolution upsampling module, the fused feature map is deconvoluted and upsampled to obtain a preprocessed feature map; Performing multi-scale feature extraction processing on the second feature map based on the feature jump module to obtain a dimensional feature map, wherein the dimensional feature map includes a first dimensional feature map, a second dimensional feature map, and a third dimensional feature map; performing interpolation processing on the dimensional feature map based on the bilinear interpolation module to obtain an interpolated dimensional feature map; Based on the adaptive skip connection module, the interpolated dimensional feature map and the preprocessed feature map are fused to obtain three fused feature maps. Based on the feature weighted fusion module, the three fusion feature maps are weightedly fused to obtain a weighted fusion feature map; Based on the decoding convolution module, the weighted fused feature map is decoded and convolved to obtain the mangrove distribution results.

Images