A method, system and terminal for identifying and monitoring mangrove populations

The DCNN-based method for mangrove population identification and ecological monitoring addresses the limitations of existing methods by providing automated, efficient, and accurate mangrove detection and change analysis, enhancing spatial and temporal monitoring capabilities.

CN118230168BActive Publication Date: 2024-12-03GUANGDONG PROVINCIAL ACADEMY OF ENVIRONMENTAL SCI
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202410423196.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-09
Publication Date
2024-12-03
Estimated Expiration
2044-04-09

AI Technical Summary

Technical Problem

Existing mangrove population identification methods are highly dependent on professional knowledge, require significant manpower and material resources, are time-consuming, prone to weather interference, and have low automation and accuracy, especially in distinguishing mangroves from other landforms.

Method used

A method utilizing a deep convolutional neural network (DCNN) for mangrove population identification and ecological monitoring, involving preprocessing of multispectral spatial image data, training a mangrove recognition model, and performing classification and area calculation with projection coordinates and grid mosaicking to analyze population and range changes.

Benefits of technology

Enables accurate, automated, and efficient identification and monitoring of mangrove populations and ecological changes, reducing reliance on human expertise and resource investment while improving accuracy and resilience to weather conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118230168B_ABST
    Figure CN118230168B_ABST
Patent Text Reader

Abstract

The present invention discloses a method, system and terminal for population identification and ecological monitoring of mangroves, the method comprising: obtaining multiple groups of multispectral spatial image data, performing preprocessing to obtain a training data set, and performing model training to obtain a mangrove identification network model; and inputting the multispectral spatial image data to be processed, outputting population classification results and mangrove range boundaries; obtaining the population area of ​​each population according to the population classification results, and obtaining the overall range area of ​​the mangroves according to the mangrove range classification results; analyzing the population area of ​​each population and the overall range area to obtain a first analysis result and a second analysis result, so as to monitor the ecological changes of the mangroves. The present invention accurately identifies and extracts population information and overall distribution range information within the mangrove area by fusing multispectral image information, so as to ensure the statistical accuracy of the area range of each population of mangroves and the total range of mangroves.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of remote sensing science and computer science technology, and in particular to a method, system, terminal and computer-readable storage medium for mangrove population identification and ecological monitoring. Background Art

[0002] Mangroves are one of the most productive marine ecosystems in tropical and subtropical coastal areas. They play an important role in purifying seawater, preventing wind and waves, fixing carbon, and maintaining biodiversity. They are important habitats for rare and endangered waterfowl, and also a place for fish, shrimp, crabs, and shellfish to grow and reproduce. Therefore, it is necessary to efficiently and accurately extract mangrove vegetation information and monitor it in real time. However, most mangroves grow in special areas such as intertidal zones and swamps. When conducting large-scale monitoring, traditional manual field survey methods require a lot of manpower, material resources, and time costs. The rapid development of drone remote sensing technology provides a powerful means for extracting and monitoring mangrove information, which can achieve high temporal and spatial resolution, fast and accurate observation.

[0003] At present, the methods for identifying mangroves mainly include manual visual interpretation, vegetation index and other methods; the manual visual interpretation method has been widely used for a period of time because it does not require on-site surveys, but it is highly dependent on the professional knowledge of the interpreters, and requires large investments in manpower and material resources. It is impossible to complete the identification of large-scale mangrove population information in a short period of time, and the degree of automation is low; the method that only relies on vegetation index to identify mangroves and their populations is easily affected by fog, clouds and rainy weather, and it is difficult to distinguish mangroves from other landforms, and the accuracy of identification is limited.

[0004] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0005] The main purpose of the present invention is to provide a method, system and terminal for mangrove population identification and ecological monitoring, aiming to solve the problem that the existing technology is highly dependent on the professional knowledge of interpreters, requires large investment in manpower and material resources, cannot complete the identification of large-scale mangrove population information in a short period of time, has a low degree of automation, is easily affected by fog, clouds and rainy weather, and has difficulty in distinguishing mangroves from other objects, resulting in low identification accuracy.

[0006] To achieve the above object, the present invention provides a method for population identification and ecological monitoring of mangroves, the method comprising the following steps:

[0007] Acquire multiple groups of multispectral spatial image data, pre-process each group of the multispectral spatial image data to obtain a training data set, and train a deep convolutional neural network training model according to the training data set to obtain a mangrove recognition network model;

[0008] Acquire the multispectral spatial image data to be processed, input the multispectral spatial image data to be processed into the mangrove identification network model, and output the population classification result and the mangrove range classification result;

[0009] Assigning projection coordinates to the population classification results to obtain target population classification results, performing grid mosaicking on the target population classification results to obtain a target grid dataset, and calculating the population area of ​​each population based on the target grid dataset, and obtaining the overall range area of ​​the mangroves based on the mangrove range classification results;

[0010] The population distribution changes of mangroves are analyzed according to the population area of ​​each population to obtain a first analysis result, the overall changes in the range of mangroves are analyzed according to the overall range area to obtain a second analysis result, and the ecological changes of the mangroves are monitored according to the first analysis result and the second analysis result to obtain a monitoring result.

[0011] Optionally, the method for population identification and ecological monitoring of mangroves, wherein the step of acquiring multiple sets of multispectral spatial image data and preprocessing each set of the multispectral spatial image data to obtain a training data set, specifically includes:

[0012] Acquire multiple sets of low-altitude high-resolution multispectral spatial image data, perform projection difference correction and image mosaic processing on each set of the multispectral spatial image data to obtain multiple corresponding orthophoto images, and perform bit depth conversion on all the orthophoto images to obtain multiple target orthophoto images;

[0013] Performing projection transformation on all the target orthophoto images to obtain a plurality of coordinate transformation images, performing image resampling on all the coordinate transformation images to obtain a plurality of target resolution images, and setting a training area;

[0014] Performing vector selection on all the target resolution images according to the training area to obtain a plurality of vector ranges, and performing cropping processing on the corresponding target resolution image according to each of the vector ranges to obtain a plurality of target raster image data;

[0015] All the target raster image data are segmented according to the degree of overlap to obtain a plurality of target segmented image data, and each target segmented image data is used as a sample group, and all the sample groups are combined to obtain a training data set for model training.

[0016] Optionally, the method for population identification and ecological monitoring of mangroves, wherein the deep convolutional neural network training model is trained according to the training data set to obtain a mangrove identification network model, specifically includes:

[0017] Creating a deep convolutional neural network training model, and inputting a sample group in the training data set into the deep convolutional neural network training model;

[0018] Performing convolution operation processing on the target segmented image data to obtain convolution features, performing attention calculation on the convolution features according to the first formula to obtain a channel attention map, and performing attention calculation on the convolution features according to the second formula to obtain a spatial attention map;

[0019] Cutting the target segmented image according to a preset ratio to obtain a plurality of multi-channel images, flattening all the multi-channel images to obtain a plurality of target vectors, and performing matrix conversion on all the target vectors to obtain a target vector matrix;

[0020] Performing information fusion on the target vector matrix according to the channel attention map and the spatial attention map to obtain target semantic information, identifying and classifying the target semantic information according to the third formula to obtain a predicted classification result of the mangrove population, and performing regional classification on the target semantic information according to the fourth formula to obtain a predicted overall mangrove area result;

[0021] Performing maximum pooling calculation on the predicted classification result to obtain predicted mangrove area boundary information, calculating the true classification result of the mangrove population corresponding to the target segmented image data and the first loss value of the predicted classification result, and calculating the true mangrove overall area result corresponding to the target segmented image data and the second loss value of the predicted mangrove overall area result;

[0022] The parameters of the deep convolutional neural network training model are modified according to the first loss value and the second loss value, and the next sample group is input into the deep convolutional neural network training model until the training status of the deep convolutional neural network training model meets the preset conditions, thereby obtaining a trained mangrove identification network model.

[0023] Optionally, in the mangrove population identification and ecological monitoring method, the first formula is:

[0024] M c (F)=δ(MLP(AvgPool(F))+MLP(MaxPool(F)));

[0025] The second formula is:

[0026] M s (F) = δ(f n*n ([AvgPool(F)];MaxPool(F)]));

[0027] The third formula is:

[0028]

[0029] The fourth formula is:

[0030]

[0031] Among them, M c (F) is the channel weight coefficient of the result of the neural network model encoding process, M s (F) is the spatial weight coefficient of the result of the neural network model encoding process, δ(z) is the Softmax function, δ'(z) is the Sigmoid function, δ is the weight coefficient of different channels obtained by using the Sigmoid activation function, MLP is a multi-layer perceptron with shared weights, AvgPool is the maximum pooling operation, MaxPool is the average pooling operation, F is the result of the neural network model encoding process, f n*n is the n*n convolution operation, z i is the output value of the i-th node, z j is the output value of the jth node, K is the number of classification categories, and z is the single original output value of the function.

[0032] Optionally, the method for population identification and ecological monitoring of mangroves, wherein the projected coordinates are assigned to the population classification results to obtain target population classification results, the target population classification results are grid mosaicked to obtain a target grid dataset, and the population area of ​​each population is calculated according to the target grid dataset, and the overall range area of ​​the mangroves is obtained according to the mangrove range classification results, specifically including:

[0033] Acquire projection coordinate information of the multispectral spatial image data to be processed, and assign projection coordinates to the population classification result using the projection coordinate information to obtain a target population classification result;

[0034] Performing raster mosaic on the classification results of the target population to obtain a target raster dataset, performing transformation processing on the target raster dataset to obtain first closed vector surface data, and performing redundant polygon elimination on the first closed vector surface data to obtain the population area of ​​each population in the mangrove forest;

[0035] The projection coordinate information is used to assign projection coordinates to the mangrove range classification result to obtain a target mangrove range classification result, the target mangrove range classification result is grid-mosaiced to obtain a target data set, the target data set is transformed to obtain second closed vector surface data, and redundant polygons are eliminated from the second closed vector surface data to obtain the overall range area of ​​the mangroves.

[0036] Optionally, the method for population identification and ecological monitoring of mangroves, wherein the step of obtaining the area of ​​a mangrove region according to the target raster dataset and the mangrove range boundary further comprises:

[0037] Obtaining the time information and mangrove area of ​​the target raster dataset, and creating a text file according to the overall area;

[0038] The time information and the mangrove area are input into the text file to obtain a target text file, data visualization is performed on the target text file to obtain visualization data, and the visualization data is displayed to obtain a display result.

[0039] Optionally, the method for identifying and monitoring the population of mangroves and their ecology, wherein the population distribution changes of mangroves are analyzed according to the population area of ​​each population to obtain a first analysis result, the overall changes in the range of mangroves are analyzed according to the overall range area to obtain a second analysis result, and the ecological changes of the mangroves are monitored according to the first analysis result and the second analysis result to obtain a monitoring result, specifically includes:

[0040] Analyzing the population distribution changes of mangroves according to the population area of ​​each population to obtain a first analysis result, and analyzing the overall changes of the mangrove range according to the overall range area to obtain a second analysis result;

[0041] Checking whether the first analysis result is a decrease in the population area of ​​mangroves, and whether the second analysis result is a decrease in the range of mangroves;

[0042] If the first analysis result is that the population area of ​​mangroves has decreased and / or the second analysis result is that the range of mangroves has decreased, then the monitoring result is to strengthen the supervision and protection of mangroves;

[0043] If the first analysis result is that the population area of ​​mangroves has increased and the second analysis result is that the range of mangroves has increased, then the monitoring result is to strengthen the ecological restoration of mangrove wetlands.

[0044] Optionally, the mangrove population identification and ecological monitoring method, wherein the mangrove population identification and ecological monitoring system comprises:

[0045] A model training module is used to obtain multiple groups of multispectral spatial image data, pre-process each group of the multispectral spatial image data to obtain a training data set, and train a deep convolutional neural network training model according to the training data set to obtain a mangrove recognition network model;

[0046] A data processing module, used for acquiring the multispectral spatial image data to be processed, inputting the multispectral spatial image data to be processed into the mangrove identification network model, and outputting population classification results and mangrove classification results;

[0047] An area calculation module is used to assign projection coordinates to the population classification results to obtain target population classification results, perform grid mosaic on the target population classification results to obtain a target grid dataset, and calculate the population area of ​​each population based on the target grid dataset, and obtain the overall range area of ​​the mangroves based on the mangrove range classification results;

[0048] The ecological monitoring module is used to analyze the changes in the population distribution of mangroves according to the population area of ​​each population to obtain a first analysis result, analyze the overall changes in the range of mangroves according to the overall range area to obtain a second analysis result, and monitor the ecological changes of the mangroves according to the first analysis result and the second analysis result to obtain a monitoring result.

[0049] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a program stored in the memory and executable on the processor, and when the program is executed by the processor, the steps of the mangrove population identification and ecological monitoring method as described above are implemented.

[0050] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a mangrove population identification and ecological monitoring program, and when the mangrove population identification and ecological monitoring program is executed by a processor, the steps of the mangrove population identification and ecological monitoring method as described above are implemented.

[0051] In the present invention, a plurality of groups of multispectral spatial image data are obtained, each group of the multispectral spatial image data is preprocessed to obtain a training data set, and a deep convolutional neural network training model is trained according to the training data set to obtain a mangrove identification network model; the multispectral spatial image data to be processed is obtained, the multispectral spatial image data to be processed is input into the mangrove identification network model, and a population classification result and a mangrove range classification result are output; the population classification result is assigned with projection coordinates to obtain a target population classification result, the target population classification result is grid mosaicked to obtain a target grid data set, and the population area of ​​each population is calculated according to the target grid data set, and the overall range area of ​​the mangrove is obtained according to the mangrove range classification result; the population distribution change of the mangrove is analyzed according to the population area of ​​each population to obtain a first analysis result, the overall change of the mangrove range is analyzed according to the overall range area to obtain a second analysis result, and the ecological changes of the mangrove are monitored according to the first analysis result and the second analysis result to obtain a monitoring result. The present invention designs a convolutional neural network model that can realize the identification and classification of mangrove populations, so that boundary extraction can be performed while classification is performed. The population information and overall distribution range information within the mangrove area are accurately identified and extracted by fusing multispectral image information. After obtaining the identification and classification results, the area range of each mangrove population is calculated to ensure the area statistical accuracy of the various mangrove population areas and the total range of mangroves. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart of a preferred embodiment of the method for population identification and ecological monitoring of mangroves in the present invention;

[0053] Figure 2 It is a schematic diagram of the specific structure of a deep convolutional neural network training model of a preferred embodiment of the present invention;

[0054] Figure 3 It is a schematic diagram of the principle of a preferred embodiment of the mangrove population identification and ecological monitoring system of the present invention;

[0055] Figure 4 It is a schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0058] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0059] The method for population identification and ecological monitoring of mangroves described in the preferred embodiment of the present invention is as follows: Figure 1 As shown, the mangrove population identification and ecological monitoring method comprises the following steps:

[0060] Step S10: Acquire multiple groups of multispectral spatial image data, pre-process each group of the multispectral spatial image data to obtain a training data set, and train a deep convolutional neural network training model according to the training data set to obtain a mangrove recognition network model.

[0061] Specifically, in recent years, deep learning has made many progresses in speech recognition, image target recognition and detection, natural language processing, etc., and image interpretation and object recognition in the field of remote sensing are essentially digital image processing. Therefore, in the present invention, DCNN (Deep Convolution Neural Network) prediction is applied to the work of mangrove population identification and change analysis in the field of ecological monitoring. The present invention introduces a deep convolutional neural network for training to obtain a model that can intelligently and accurately extract deep semantic information of offshore mangroves and their spectral radiation in high-resolution unmanned aerial vehicle multispectral spatial data, namely, a mangrove recognition network model, wherein the mangrove recognition network model is obtained by model training based on a training data set, and the process of obtaining the training data set is as follows: first, a plurality of sets of low-altitude high-resolution multispectral spatial image data are obtained by using an unmanned aerial vehicle, and all multispectral spatial image data are preprocessed, including orthophoto generation, bit depth conversion, UTM projection conversion, image resampling, training data set selection, vector range clipping and overlap segmentation.

[0062] The processing process of orthophoto generation is to generate corresponding orthophoto maps for each single band (for example, R (Red, visible light-red), RE (Red Edge, red edge band), G (Green, visible light-green), B (Blue, visible light-blue), NIR (Near-Infrared, near infrared) and SWIR (Short Wave Infrared, short wave infrared) spectral bands) in each group of multispectral spatial image data, and obtain multiple corresponding orthophoto maps. Because ordinary drone photos only have the coordinates of the center point of the photo, but the photo has no projection coordinate system, and each pixel has no projection coordinate representation, by establishing orthophotos, one is to stitch drone photos on a large scale, and the other is to set specific projection coordinates for the image and each pixel, so as to facilitate subsequent spatial calculation and analysis; the processing process of bit depth conversion is that the bit depth of the image in the multispectral spatial image data obtained by the drone is 16-B it, in order to facilitate visualization operation, in the embodiment of the present invention, the image with a bit depth of 16-bit is converted to an image with a bit depth of 8-bit; the gray value of the 16-bit multispectral spatial image data is linearly stretched to 8-bit through the bit depth conversion module, which is convenient for the subsequent training and prediction of the deep convolutional neural network model, that is, all the orthophotos are converted to bit depth to obtain multiple target orthophotos; the processing process of the UTM projection conversion is that in order to facilitate the grid mosaic work of the obtained prediction results, the geographic latitude and longitude coordinate system WGS (World Geographic System) needs to be converted to the grid mosaic work. The image resampling process is as follows: in order to ensure that the images of different series of training data sets have consistent resolution, the multispectral spatial image data of different bands need to be unified to the same resolution size to ensure that the training data size of each band of UAV remote sensing images is consistent, that is, all the coordinate transformation images are resampled to obtain multiple target resolution images; the training data set selection process is as follows: after completing the image resampling, some mangrove wetland ecological areas need to be selected for model training. In the selection process, the principles of uniform distribution and comprehensive samples should be followed, that is, the training area should be set;The processing process of the vector range clipping is to select the area of ​​each band suitable for model training by vector box selection, and use the vector range to clip the target area in the image, that is, the raster area covered by the vector Shapefile data will be clipped separately to form raster image data consistent with the area range of the vector Shapefile, that is, all the target resolution images are vector box selected according to the training area to obtain multiple vector ranges, and the corresponding target resolution images are clipped according to each vector range to obtain multiple target raster image data; the processing process of the overlap segmentation is to segment the clipping result with a certain overlap, and the overlap segmentation It can effectively ensure the sufficiency of the training set data volume, and at the same time, it can make full use of the image edge information. In the segmentation process, the projection coordinate information of the segmented image is first obtained to obtain the coordinate position of the upper left corner of the image to be segmented and the pixel unit resolution size, and according to the overlapping range of the segmented pixels, the pixel interval of the next segmented image from the previous segmented image is calculated, so as to calculate the starting position of the segmentation at the next moment, and use the starting position as the coordinate of the upper left corner of the next segmented image, and finally obtain each independent segmented image, that is, the target segmented image, according to the segmentation length and width; then, each target segmented image data is taken as a sample group, and all the sample groups are combined to obtain the training data set for model training. ;

[0063] After obtaining the training data set for model training, it is necessary to train the deep convolutional neural network training model according to the training data set to obtain the mangrove identification network model. Specifically, first, a deep convolutional neural network training model is created. The specific structure of the deep convolutional neural network training model is as follows: Figure 2As shown, then, a sample group in the training data set is input into the deep convolutional neural network training model; the convolution module in the deep convolutional neural network training model performs convolution operation on the target segmented image data to obtain convolution features. The convolution module uses the advantages of convolutional neural networks in image processing and a series of convolution sliding calculation methods to perform convolution operations on the input image according to the convolution kernel size and the sliding step size, thereby obtaining deep semantic information of the input image. The convolution module not only sets up several convolution layers to extract deep information from remote sensing images, but also can perform operations on the image information extracted from the previous layer and the current convolution feature layer, making fuller use of shallow and deep information, thereby improving operation efficiency. Afterwards, in the process of performing convolution operations on the image, the Spatial Attention mechanism and the Channel are added before the Double Attention module in the deep convolutional neural network training model performs deconvolution. Attention (channel attention) mechanism. In the dual attention module, the spatial attention mechanism and the channel attention mechanism are set in parallel. When performing channel attention, the weight of each channel can be learned to generate an attention map in the channel domain, that is, the channel attention map. The specific calculation formula is:

[0064] M c (F)=δ(MLP(AvgPool(F))+MLP(MaxPool(F)));

[0065] Among them, M c (F) is the channel weight coefficient of the result of the neural network model encoding process, δ is the weight coefficient of different channels obtained by using the Sigmoid activation function, MLP is a multi-layer perceptron with shared weights, AvgPool is the maximum pooling operation, MaxPool is the average pooling operation, and F is the result of the neural network model encoding process; the spatial attention mechanism is different from the channel attention mechanism. The spatial attention mechanism mainly focuses on the context position information. First, MaxPooling (maximum pooling operation) and AveragePooling (average pooling operation) are used on the channel dimension to obtain two different feature layers, and then the two feature layers are merged, and the convolution operation is used to generate the Spatial Attention Map. The specific calculation formula is:

[0066] M s (F) = δ(f n*n ([AvgPool(F)];MaxPool(F)]));

[0067] Among them, M s(F) is the spatial weight coefficient of the result of the neural network model encoding process, f n*n It is an n*n convolution operation.

[0068] The Vision Transformer (visual self-attention) module of the deep convolutional neural network training model focuses on the acquisition of global information related features, that is, the acquisition of information in the entire image or the entire data set, and the global information usually requires analysis and learning of the entire data set to obtain global patterns and laws; in the present invention, the visual self-attention module first cuts each mangrove multispectral image, and divides the input 512*512 image according to a preset ratio, such as 16*16, to obtain a total of 1024 Patches after division. Patch is a small multi-channel image, that is, a multi-channel image. The number of channels is equal to the number of spectra, and the style of Patch is (16, 16, n). Flattening it becomes a vector with a length of 16*16*n, namely the target vector. Then, all the target vectors are converted into a matrix of [1024, 16*16*n], namely the target vector matrix, for parallel input. Then, the target vector matrix is ​​fused according to the channel attention map and the spatial attention map to obtain the target semantic information. Then, the Softmax classification module and the Sigmoid classification module of the deep convolutional neural network training model need to be used to predict the results.

[0069] In the Softmax classification module, the Softmax function is used to automatically identify and classify the target semantic information extracted through a series of convolutions to obtain the predicted classification results of the mangrove population. The expression of the Softmax function is:

[0070]

[0071] Among them, δ(z) is the Softmax function, z i is the output value of the i-th node, z j is the output value of the jth node, K is the number of classification categories; when the Softmax function tends to positive infinity or negative infinity, the function approaches a smooth state, and because the output range of the Softmax function is (0, 1), and the sum of each output value is 1, the Softmax function is selected in the Softmax classification module to perform probability statistics of multiple classifications (various mangrove populations) to obtain the predicted classification results of the mangrove population.

[0072] In the Sigmoid classification module, the Sigmoid function is used to automatically identify and classify the target semantic information extracted through a series of convolutions to obtain the predicted overall mangrove area result, including mangrove areas and non-mangrove areas; the expression of the Sigmoid function is:

[0073]

[0074] Among them, δ'(z) is the Sigmoid function, and z is the single original output value of the function; when the Sigmoid function tends to positive infinity or negative infinity, the function approaches a smooth state, and because the output range of the Sigmoid function is (0, 1), this function is selected in the Sigmoid classification module to perform probability statistics of binary classification (mangrove area and non-mangrove area) to obtain the predicted results of the overall mangrove area.

[0075] After obtaining the predicted mangrove overall area result, the predicted mangrove overall area result is subjected to the maximum pooling (MaxPooling) calculation by the boundary extraction module of the deep convolutional neural network training model, and the mangrove area range boundary information can be obtained at the same time by subtracting the original Sigmoid result layer, that is, the predicted mangrove area boundary information; then, it is necessary to calculate the true classification result of the mangrove population corresponding to the target segmented image data and the first loss value of the predicted classification result, and calculate the true mangrove overall area result corresponding to the target segmented image data and the second loss value of the predicted mangrove overall area result, in order to obtain the mangrove area boundary information according to the first The loss value and the second loss value are used to correct the parameters of the deep convolutional neural network training model, and the next sample group is input into the deep convolutional neural network training model, and the above process is repeated, which will not be repeated here; until the training status of the deep convolutional neural network training model meets the preset conditions, the preset conditions include that the target loss value meets the preset requirements or the number of training times reaches the preset number, the preset requirements can be determined according to the accuracy of the mangrove identification network model, which will not be described in detail here, the preset number can be the maximum number of training times of the deep convolutional neural network training model, for example, 2000 times, etc., finally to obtain the trained mangrove identification network model.

[0076] Step S20, obtaining the multispectral spatial image data to be processed, inputting the multispectral spatial image data to be processed into the mangrove identification network model, and outputting the population classification result and the mangrove range boundary.

[0077] Specifically, after obtaining the trained mangrove identification network model, the multispectral spatial image data to be processed is obtained, and the multispectral spatial image data to be processed is input into the mangrove identification network model, and the mangrove identification network model directly outputs the corresponding population classification results and mangrove range classification results.

[0078] Step S30, assign projection coordinates to the population classification results to obtain target population classification results, perform grid mosaicking on the target population classification results to obtain a target grid dataset, and calculate the population area of ​​each population based on the target grid dataset, and obtain the overall range area of ​​the mangroves based on the mangrove range classification results.

[0079] The step S30 comprises:

[0080] Step S31, obtaining the projection coordinate information of the multispectral spatial image data to be processed, assigning the projection coordinate information to the population classification result, and obtaining the target population classification result;

[0081] Step S32, performing grid mosaic on the target population classification result to obtain a target grid data set, converting the target grid data set to obtain first closed vector surface data, and removing redundant polygons from the first closed vector surface data to obtain the population area of ​​each population in the mangrove forest;

[0082] Step S33, assigning projection coordinates to the mangrove range classification result using the projection coordinate information to obtain a target mangrove range classification result, performing grid mosaicking on the target mangrove range classification result to obtain a target data set, transforming the target data set to obtain second closed vector surface data, and removing redundant polygons from the second closed vector surface data to obtain the overall range area of ​​the mangroves.

[0083] Specifically, the population classification results and mangrove range classification results output by the mangrove identification network model are sequentially read into the post-processing program, and the GDAL (Geospatial Data Abstraction Library) Python API (Application Programming Interface) is used to read the corresponding data into an array form for subsequent processing; when the population classification result is obtained, the population classification result has been segmented and does not have geographic coordinates, and it is necessary to add projection coordinate information to the population classification result for subsequent raster mosaic work; first, the projection coordinate information of the original segmented data that has not been processed by the mangrove identification network model, that is, the projection coordinate information of the multispectral spatial image data to be processed, and then the projection coordinate information is assigned to the corresponding population classification result, and a new result image data with projection coordinate information is generated, that is, the target population classification result. In addition, when executing the program, it is necessary to release the image that has been assigned last time to avoid occupying the internal processor memory space; then, using ArcGIS Pro's ArcPy interface, and according to the target population classification results, it can be mosaicked into a large-scale raster dataset. The specific implementation process includes reading each segmented raster image, specifying the bit depth, output coordinate system and stitching algorithm for its mosaicked result, and finally obtaining the target raster dataset; when the mangrove range classification result is obtained, this mangrove range classification result has been segmented and does not have geographic coordinates. It is necessary to add projection coordinate information to this mangrove range classification result for subsequent raster mosaic work; first, obtain the projection coordinate information of the original segmented data that has not been processed by the mangrove identification network model, that is, the projection coordinate information of the multispectral spatial image data to be processed, and then assign the projection coordinate information to the corresponding mangrove range classification result, and generate new result image data with projection coordinate information, that is, the target mangrove range classification result; then, using ArcGIS Pro's ArcPy interface, according to the target mangrove range classification result, it can be mosaicked into a large-scale raster dataset. The specific implementation process includes reading each segmented raster image, specifying the bit depth, output coordinate system and stitching algorithm for its mosaicked result, and finally obtaining the target dataset;After that, it is necessary to calculate the population area of ​​each population in the corresponding mangrove forest and the overall range area of ​​the mangrove forest. In an embodiment of the present invention, two calculation methods are provided, namely, area calculation based on pixel statistics and area calculation based on an appropriate amount of boundary closed area. The process of area calculation based on pixel statistics is to count the number of pixels of each population within the scope of the target population classification result, and calculate the population area of ​​each population in the mangrove forest and the overall range area of ​​the mangrove forest according to the resolution of the multispectral spatial image data to be processed, and convert the final result from square meters to square kilometers to obtain the population area of ​​each population in the mangrove forest and the overall range area of ​​the mangrove forest; The process of area calculation based on the appropriate amount of boundary closed area is to convert the extracted corresponding data set into closed vector surface data, then remove redundant polygons from the vector surface data, and then calculate the vector range area, so as to obtain the population area of ​​each population in the mangrove forest and the overall range area of ​​the mangrove forest, that is, to convert the target raster data set to obtain the first closed vector surface data, remove redundant polygons from the first closed vector surface data, and obtain the population area of ​​each population in the mangrove forest; to convert the target raster data set to obtain the second closed vector surface data, remove redundant polygons from the second closed vector surface data, and obtain the overall range area of ​​the mangrove forest. ;

[0084] Furthermore, the time information of the target raster dataset and the target dataset is obtained, and the time information is converted into a time format. Specifically, the file name character string representing the time is converted into a time format recognizable in the subsequent visualization function, for example, "20231215" is converted to "2023.95" to facilitate subsequent visualization analysis; then, the time and mangrove range area corresponding to each target raster dataset and the target dataset are output to the created text file to obtain the target text file; then, the target text file is visualized to facilitate intuitive analysis of the interannual and seasonal changes in the area of ​​various mangrove populations, and the boundary raster data of the mangrove area is converted into a vector format, and overlaid analysis is performed in chronological order to obtain intuitive changes in the range and boundaries of the mangrove area in the time series.

[0085] Step S40, analyzing the population distribution changes of mangroves according to the population area of ​​each population to obtain a first analysis result, analyzing the overall changes in the range of mangroves according to the overall range area to obtain a second analysis result, and monitoring the ecological changes of the mangroves according to the first analysis result and the second analysis result to obtain a monitoring result.

[0086] The step S40 comprises:

[0087] Step S41, analyzing the population distribution changes of mangroves according to the population area of ​​each population to obtain a first analysis result, and analyzing the overall changes of the mangrove range according to the overall range area to obtain a second analysis result;

[0088] Step S42, checking whether the first analysis result is that the population area of ​​mangroves has decreased, and whether the second analysis result is that the range of mangroves has decreased;

[0089] Step S43: If the first analysis result is that the population area of ​​mangroves has decreased and / or the second analysis result is that the range of mangroves has decreased, then the monitoring result is to strengthen the supervision and protection of mangroves;

[0090] Step S44: If the first analysis result is that the population area of ​​mangroves has increased and the second analysis result is that the range of mangroves has increased, then the monitoring result is to strengthen the ecological restoration of mangrove wetlands.

[0091] Specifically, the changes in the population distribution of mangroves are analyzed according to the population area of ​​each population to obtain a first analysis result, and the overall changes in the range of mangroves are analyzed according to the overall range area to obtain a second analysis result, wherein the first analysis result and the second analysis result are both visualization analysis results of spatiotemporal changes, and then the ecological protection and restoration evaluation of offshore mangrove wetland ecosystems can be carried out based on this analysis result; the ecological changes of the mangroves are monitored according to this analysis result, if the first analysis result is a decrease in the population area of ​​mangroves and the second analysis result is a decrease in the range of mangroves, the first analysis result is a decrease in the population area of ​​mangroves and the second analysis result is an increase in the range of mangroves, and the first analysis result is an increase in the population area of ​​mangroves and the second analysis result is a decrease in the range of mangroves, then the monitoring result is to strengthen the supervision and protection of mangroves; if the first analysis result is an increase in the population area of ​​mangroves and the second analysis result is an increase in the range of mangroves, then the monitoring result is to strengthen the ecological restoration of mangrove wetlands.

[0092] The present invention trains a deep convolutional neural network training model. The trained model can realize mangrove population identification and classification, and can also perform boundary extraction. Then, when performing data processing, the recognition method of ordinary single-band or RGB images is improved, and multi-spectral image information is integrated to accurately identify and extract the population information and overall distribution range information within the mangrove area. After obtaining the identification and classification results, a pixel-based statistics method is used to calculate the area range of each mangrove population, thereby ensuring the accuracy of the area statistics of the various mangrove population areas and the total range of mangroves.

[0093] Furthermore, if Figure 3As shown, based on the above-mentioned mangrove population identification and ecological monitoring method, the present invention also provides a mangrove population identification and ecological monitoring system, and the mangrove population identification and ecological monitoring system includes:

[0094] A model training module 51 is used to obtain multiple groups of multispectral spatial image data, pre-process each group of the multispectral spatial image data to obtain a training data set, and train a deep convolutional neural network training model according to the training data set to obtain a mangrove recognition network model;

[0095] The data processing module 52 is used to obtain the multi-spectral spatial image data to be processed, input the multi-spectral spatial image data to be processed into the mangrove identification network model, and output the population classification result and the mangrove range classification result;

[0096] The area calculation module 53 is used to assign projection coordinates to the population classification results to obtain target population classification results, perform grid mosaic on the target population classification results to obtain a target grid dataset, and calculate the population area of ​​each population according to the target grid dataset, and obtain the overall range area of ​​the mangroves according to the mangrove range classification results;

[0097] The ecological monitoring module 54 is used to analyze the changes in the population distribution of mangroves according to the population area of ​​each population to obtain a first analysis result, analyze the overall changes in the range of mangroves according to the overall range area to obtain a second analysis result, and monitor the ecological changes of the mangroves according to the first analysis result and the second analysis result to obtain a monitoring result.

[0098] Furthermore, if Figure 4 As shown, based on the above-mentioned mangrove population identification and ecological monitoring method, the present invention also provides a terminal accordingly, and the terminal includes a processor 10, a memory 20 and a display 30. Figure 4 Only some components of the terminal are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0099] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a mangrove population identification and ecological monitoring program 40 is stored on the memory 20, and the mangrove population identification and ecological monitoring program 40 can be executed by the processor 10, thereby realizing the mangrove population identification and ecological monitoring method in the present application.

[0100] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing the mangrove population identification and ecological monitoring method.

[0101] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0102] In one embodiment, when the processor 10 executes the program 40 for mangrove species identification and ecological monitoring in the memory 20 , the steps of the mangrove species identification and ecological monitoring method described above are implemented.

[0103] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a mangrove population identification and ecological monitoring program, and when the mangrove population identification and ecological monitoring program is executed by a processor, the steps of the mangrove population identification and ecological monitoring method as described above are implemented.

[0104] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0105] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium that can be read by a computer, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0106] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for population identification and ecological monitoring of mangroves, characterized in that: The mangrove population identification and ecological monitoring method includes: Acquire multiple groups of multispectral spatial image data, pre-process each group of the multispectral spatial image data to obtain a training data set, and train a deep convolutional neural network training model according to the training data set to obtain a mangrove recognition network model; The acquiring of multiple groups of multispectral spatial image data and preprocessing of each group of the multispectral spatial image data to obtain a training data set specifically includes: Acquire multiple sets of low-altitude high-resolution multispectral spatial image data, perform projection difference correction and image mosaic processing on each set of the multispectral spatial image data to obtain multiple corresponding orthophoto images, and perform bit depth conversion on all the orthophoto images to obtain multiple target orthophoto images; Performing projection transformation on all the target orthophoto images to obtain a plurality of coordinate transformation images, performing image resampling on all the coordinate transformation images to obtain a plurality of target resolution images, and setting a training area; Performing vector selection on all the target resolution images according to the training area to obtain a plurality of vector ranges, and performing cropping processing on the corresponding target resolution image according to each of the vector ranges to obtain a plurality of target raster image data; Performing overlapping segmentation on all the target raster image data to obtain a plurality of target segmented image data, and taking each target segmented image data as a sample group, and combining all the sample groups to obtain a training data set for model training; The deep convolutional neural network training model is trained according to the training data set to obtain a mangrove identification network model, specifically including: Creating a deep convolutional neural network training model, and inputting a sample group in the training data set into the deep convolutional neural network training model; Performing convolution operation processing on the target segmented image data to obtain convolution features, performing attention calculation on the convolution features according to the first formula to obtain a channel attention map, and performing attention calculation on the convolution features according to the second formula to obtain a spatial attention map; Cutting the target segmented image according to a preset ratio to obtain a plurality of multi-channel images, flattening all the multi-channel images to obtain a plurality of target vectors, and performing matrix conversion on all the target vectors to obtain a target vector matrix; Performing information fusion on the target vector matrix according to the channel attention map and the spatial attention map to obtain target semantic information, identifying and classifying the target semantic information according to the third formula to obtain a predicted classification result of the mangrove population, and performing regional classification on the target semantic information according to the fourth formula to obtain a predicted overall mangrove area result; Performing maximum pooling calculation on the predicted classification result to obtain predicted mangrove area boundary information, calculating the true classification result of the mangrove population corresponding to the target segmented image data and the first loss value of the predicted classification result, and calculating the true mangrove overall area result corresponding to the target segmented image data and the second loss value of the predicted mangrove overall area result; Modifying the parameters of the deep convolutional neural network training model according to the first loss value and the second loss value, and inputting the next sample group into the deep convolutional neural network training model until the training status of the deep convolutional neural network training model meets the preset conditions, thereby obtaining a trained mangrove identification network model; The first formula is: M c (F)=δ(MLP(AvgPool(F))+MLP(MaxPool(F))); The second formula is: M s (F)=δ(f n*n ([AvgPool(F)];[MaxPool(F)])); The third formula is: The fourth formula is: Among them, M c (F) is the channel weight coefficient of the result of the neural network model encoding process, M s (F) is the spatial weight coefficient of the result of the neural network model encoding process, δ(z) is the Softmax function, δ ′ (z) is the Sigmoid function, δ is the weight coefficient of different channels obtained by using the Sigmoid activation function, MLP is a multi-layer perceptron with shared weights, AvgPool is the average pooling operation, MaxPool is the maximum pooling operation, F is the result of the neural network model encoding process, and f n*n is the n*n convolution operation, z i is the output value of the i-th node, z j is the output value of the jth node, K is the number of categories, and z is the single original output value of the function; Acquire the multispectral spatial image data to be processed, input the multispectral spatial image data to be processed into the mangrove identification network model, and output the population classification result and the mangrove range classification result; Assigning projection coordinates to the population classification results to obtain target population classification results, performing grid mosaicking on the target population classification results to obtain a target grid dataset, and calculating the population area of ​​each population based on the target grid dataset, and obtaining the overall range area of ​​the mangroves based on the mangrove range classification results; The projected coordinates are assigned to the population classification results to obtain target population classification results, the target population classification results are grid mosaicked to obtain a target grid dataset, and the population area of ​​each population is calculated according to the target grid dataset, and the overall range area of ​​the mangroves is obtained according to the mangrove range classification results, specifically including: Acquire projection coordinate information of the multispectral spatial image data to be processed, and assign projection coordinates to the population classification result using the projection coordinate information to obtain a target population classification result; Performing raster mosaic on the classification results of the target population to obtain a target raster dataset, performing transformation processing on the target raster dataset to obtain first closed vector surface data, and performing redundant polygon elimination on the first closed vector surface data to obtain the population area of ​​each population in the mangrove forest; Assigning projection coordinates to the mangrove range classification result using the projection coordinate information to obtain a target mangrove range classification result, performing grid mosaic on the target mangrove range classification result to obtain a target data set, performing transformation processing on the target data set to obtain second closed vector surface data, performing redundant polygon elimination on the second closed vector surface data to obtain the overall range area of ​​the mangrove; The population distribution changes of mangroves are analyzed according to the population area of ​​each population to obtain a first analysis result, the overall changes in the range of mangroves are analyzed according to the overall range area to obtain a second analysis result, and the ecological changes of the mangroves are monitored according to the first analysis result and the second analysis result to obtain a monitoring result.

2. The method for mangrove population identification and ecological monitoring according to claim 1, characterized in that: The method further comprises: performing projection coordinate assignment on the population classification result to obtain a target population classification result, performing grid mosaic on the target population classification result to obtain a target grid dataset, and calculating the population area of ​​each population according to the target grid dataset, and obtaining the overall range area of ​​the mangroves according to the mangrove range classification result, and then further comprising: Obtaining the time information and mangrove area of ​​the target raster dataset, and creating a text file according to the overall area; The time information and the mangrove area are input into the text file to obtain a target text file, data visualization is performed on the target text file to obtain visualization data, and the visualization data is displayed to obtain a display result.

3. The method for mangrove population identification and ecological monitoring according to claim 1, characterized in that: The first analysis result is obtained by analyzing the population distribution change of the mangroves according to the population area of ​​each population, the second analysis result is obtained by analyzing the overall change of the mangrove range according to the overall range area, and the ecological change of the mangroves is monitored according to the first analysis result and the second analysis result to obtain the monitoring result, which specifically includes: Analyzing the population distribution changes of mangroves according to the population area of ​​each population to obtain a first analysis result, and analyzing the overall changes of the mangrove range according to the overall range area to obtain a second analysis result; Checking whether the first analysis result is a decrease in the population area of ​​mangroves, and whether the second analysis result is a decrease in the range of mangroves; If the first analysis result is that the population area of ​​mangroves has decreased and / or the second analysis result is that the range of mangroves has decreased, then the monitoring result is to strengthen the supervision and protection of mangroves; If the first analysis result is that the population area of ​​mangroves has increased and the second analysis result is that the range of mangroves has increased, then the monitoring result is to strengthen the ecological restoration of mangrove wetlands.

4. A mangrove population identification and ecological monitoring system, characterized in that: The mangrove population identification and ecological monitoring system includes: A model training module is used to obtain multiple groups of multispectral spatial image data, pre-process each group of the multispectral spatial image data to obtain a training data set, and train a deep convolutional neural network training model according to the training data set to obtain a mangrove recognition network model; The acquiring of multiple groups of multispectral spatial image data and preprocessing of each group of the multispectral spatial image data to obtain a training data set specifically includes: Acquire multiple sets of low-altitude high-resolution multispectral spatial image data, perform projection difference correction and image mosaic processing on each set of the multispectral spatial image data to obtain multiple corresponding orthophoto images, and perform bit depth conversion on all the orthophoto images to obtain multiple target orthophoto images; Performing projection transformation on all the target orthophoto images to obtain a plurality of coordinate transformation images, performing image resampling on all the coordinate transformation images to obtain a plurality of target resolution images, and setting a training area; Performing vector selection on all the target resolution images according to the training area to obtain a plurality of vector ranges, and performing cropping processing on the corresponding target resolution image according to each of the vector ranges to obtain a plurality of target raster image data; Performing overlapping segmentation on all the target raster image data to obtain a plurality of target segmented image data, and taking each target segmented image data as a sample group, and combining all the sample groups to obtain a training data set for model training; The deep convolutional neural network training model is trained according to the training data set to obtain a mangrove identification network model, specifically including: Creating a deep convolutional neural network training model, and inputting a sample group in the training data set into the deep convolutional neural network training model; Performing convolution operation processing on the target segmented image data to obtain convolution features, performing attention calculation on the convolution features according to the first formula to obtain a channel attention map, and performing attention calculation on the convolution features according to the second formula to obtain a spatial attention map; Cutting the target segmented image according to a preset ratio to obtain a plurality of multi-channel images, flattening all the multi-channel images to obtain a plurality of target vectors, and performing matrix conversion on all the target vectors to obtain a target vector matrix; Performing information fusion on the target vector matrix according to the channel attention map and the spatial attention map to obtain target semantic information, identifying and classifying the target semantic information according to the third formula to obtain a predicted classification result of the mangrove population, and performing regional classification on the target semantic information according to the fourth formula to obtain a predicted overall mangrove area result; Performing maximum pooling calculation on the predicted classification result to obtain predicted mangrove area boundary information, calculating the true classification result of the mangrove population corresponding to the target segmented image data and the first loss value of the predicted classification result, and calculating the true mangrove overall area result corresponding to the target segmented image data and the second loss value of the predicted mangrove overall area result; Modifying the parameters of the deep convolutional neural network training model according to the first loss value and the second loss value, and inputting the next sample group into the deep convolutional neural network training model until the training status of the deep convolutional neural network training model meets the preset conditions, thereby obtaining a trained mangrove identification network model; The first formula is: M c (F)=δ(MLP(AvgPool(F))+MLP(maxPool(F))); The second formula is: M s (F)=δ(f n*n ([AvgPool(F)];[MaxPool(F)])); The third formula is: The fourth formula is: Among them, M c (F) is the channel weight coefficient of the result of the neural network model encoding process, M s (F) is the spatial weight coefficient of the result of the neural network model encoding process, δ(z) is the Softmax function, δ ′ (z) is the Sigmoid function, δ is the weight coefficient of different channels obtained by using the Sigmoid activation function, MLP is a multi-layer perceptron with shared weights, AvgPool is the average pooling operation, MaxPool is the maximum pooling operation, F is the result of the neural network model encoding process, and f n*n is the n*n convolution operation, z i is the output value of the i-th node, z j is the output value of the jth node, K is the number of categories, and z is the single original output value of the function; A data processing module, used for acquiring the multispectral spatial image data to be processed, inputting the multispectral spatial image data to be processed into the mangrove identification network model, and outputting population classification results and mangrove classification results; An area calculation module is used to assign projection coordinates to the population classification results to obtain target population classification results, perform grid mosaic on the target population classification results to obtain a target grid dataset, and calculate the population area of ​​each population based on the target grid dataset, and obtain the overall range area of ​​the mangroves based on the mangrove range classification results; The projected coordinates are assigned to the population classification results to obtain target population classification results, the target population classification results are grid mosaicked to obtain a target grid dataset, and the population area of ​​each population is calculated according to the target grid dataset, and the overall range area of ​​the mangroves is obtained according to the mangrove range classification results, specifically including: Acquire projection coordinate information of the multispectral spatial image data to be processed, and assign projection coordinates to the population classification result using the projection coordinate information to obtain a target population classification result; Performing raster mosaic on the classification results of the target population to obtain a target raster dataset, performing transformation processing on the target raster dataset to obtain first closed vector surface data, and performing redundant polygon elimination on the first closed vector surface data to obtain the population area of ​​each population in the mangrove forest; Assigning projection coordinates to the mangrove range classification result using the projection coordinate information to obtain a target mangrove range classification result, performing grid mosaic on the target mangrove range classification result to obtain a target data set, performing transformation processing on the target data set to obtain second closed vector surface data, performing redundant polygon elimination on the second closed vector surface data to obtain the overall range area of ​​the mangrove; The ecological monitoring module is used to analyze the changes in the population distribution of mangroves according to the population area of ​​each population to obtain a first analysis result, analyze the overall changes in the range of mangroves according to the overall range area to obtain a second analysis result, and monitor the ecological changes of the mangroves according to the first analysis result and the second analysis result to obtain a monitoring result.

5. A terminal, characterized in that: The terminal includes a memory, a processor, and a program stored in the memory and executable on the processor. When the program is executed by the processor, the steps of the mangrove population identification and ecological monitoring method as described in any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium, characterized in that: A computer program is stored thereon, wherein the computer-readable storage medium stores a mangrove population identification and ecological monitoring program, and when the mangrove population identification and ecological monitoring program is executed by a processor, the steps of the mangrove population identification and ecological monitoring method as described in any one of claims 1-3 are implemented.

Citation Information

Patent Citations

  • Water body change detection method and system based on deep convolutional neural network

    CN113269028A

  • Mangrove forest hyperspectral image classification method and device, electronic equipment and storage medium

    CN113420640A

  • Mangrove forest net primary productivity estimation method based on remote sensing and process coupling model

    CN113516362A

  • Mangrove forest dynamic monitoring method fusing multiple change detection algorithms

    CN115512223A

  • Method for inhibiting scouring in front of seawall based on mangrove forest population spatial configuration

    CN115575606A