A method for predicting mangrove carbon reserves, electronic equipment and storage medium

Abstract

The invention relates to a mangrove carbon reserve prediction method, an electronic device and a storage medium, and belongs to the field of computer vision and remote sensing technology. The mangrove carbon reserve prediction method comprises the following steps: obtaining remote sensing data and measured data of mangroves, annotating the remote sensing data based on the measured data to construct a mangrove species data set; training the constructed MFPA-YOLO semi-supervised model based on the mangrove species data set, wherein the MFPA-YOLO semi-supervised model comprises an input layer, a target detection layer and an output layer, the target detection layer comprises a teacher network, a pseudo-label generation network and a student network, predicting a mangrove remote sensing image to be predicted based on the well-trained MFPA-YOLO semi-supervised model, obtaining an initial predicted value of mangrove carbon reserve, determining an ecological environment index, and determining a predicted value of mangrove carbon reserve based on the ecological environment index and the initial predicted value of mangrove carbon reserve, thereby improving the efficiency and accuracy of the model.

Description

Technical Field

[0001] The present invention relates to the field of computer vision and remote sensing technology, and in particular to a mangrove carbon reserve prediction method, electronic equipment and storage medium. Background Art

[0002] Blue carbon has the advantages of great ecological and environmental benefits, high carbon fixation rate and long-term carbon fixation capacity. In particular, mangroves have a particularly strong carbon storage and fixation capacity, and their carbon sink capacity is 50 times that of tropical rainforests. Therefore, the development of the blue carbon economy plays an important role in mitigating the greenhouse effect.

[0003] Using deep learning models to calculate mangrove carbon reserves is an effective method to improve the accuracy and efficiency of vegetation biomass and carbon sink estimation. It can provide technical support for ecological carbon benefit assessment and accounting, and also provide a scientific basis for data decision-making in climate change policy formulation and implementation management.

[0004] However, common deep learning models require a lot of manual annotation of remote sensing data when calculating mangrove carbon reserves, which not only consumes a lot of manpower and material resources for field measurements, resulting in increased costs, but also has low efficiency and accuracy of the model. Summary of the invention

[0005] In view of this, it is necessary to provide a mangrove carbon storage prediction method, electronic equipment and storage medium to solve the technical problems of low efficiency and accuracy of the model when calculating mangrove carbon storage.

[0006] In order to solve the above problems, the present invention provides a method for predicting mangrove carbon reserves, comprising:

[0007] Acquiring remote sensing data and measured data of mangroves, and annotating the remote sensing data based on the measured data to construct a mangrove species dataset, wherein the acquiring remote sensing data and measured data of mangroves, and annotating the remote sensing data based on the measured data to construct a mangrove species dataset includes:

[0008] Calculate the total aboveground biomass of mangroves based on the measured data, and determine the initial total carbon storage of mangroves based on the total aboveground biomass of mangroves, wherein the measured data include mangrove species name, breast diameter, tree height and geographic coordinates of the sample plot;

[0009] The remote sensing data are labeled based on the mangrove species name, the sample plot geographic coordinates, and the initial total carbon stock of the mangrove to construct a mangrove species dataset, wherein the mangrove species dataset includes a labeled remote sensing dataset and an unlabeled remote sensing dataset;

[0010] The constructed MFPA-YOLO semi-supervised model is trained based on the mangrove species dataset, wherein the MFPA-YOLO semi-supervised model includes an input layer, a target detection layer and an output layer, the target detection layer includes a teacher network, a pseudo-label generation network and a student network, and the constructed MFPA-YOLO semi-supervised model is trained based on the mangrove species dataset, including:

[0011] Inputting the mangrove species dataset into the MFPA-YOLO semi-supervised model, training the teacher network based on the labeled remote sensing dataset, and updating the parameters of the teacher network to obtain a fully trained teacher network;

[0012] Predicting the unlabeled remote sensing dataset based on the fully trained teacher network to obtain a pseudo-labeled dataset;

[0013] After classifying the pseudo-label data set based on the pseudo-label generation network, iteratively training the student network based on the classified pseudo-label data set and the unlabeled remote sensing data set, and updating the parameters of the student network based on the parameters of the teacher network to obtain a fully trained MFPA-YOLO semi-supervised model;

[0014] Based on the well-trained MFPA-YOLO semi-supervised model, the remote sensing images of mangroves to be predicted are predicted to obtain the initial predicted value of mangrove carbon storage, determine the ecological environmental indicators, and determine the predicted value of mangrove carbon storage based on the ecological environmental indicators and the initial predicted value of mangrove carbon storage.

[0015] In a possible implementation, the calculation formula for the total amount of above-ground organisms in mangroves is:

[0016] ,

[0017] in, is the total aboveground biomass of mangroves, is the name of the mangrove species. The first The diameter of a tree at breast height, The first The height of the trees, , , is the coefficient of species;

[0018] The calculation formula for the initial total carbon storage of mangroves is:

[0019] ,

[0020] in, is the initial total carbon storage of mangroves.

[0021] In a possible implementation, the teacher network has the same structure as the student network, the teacher network includes a backbone network layer, a neck fusion layer and a head detection layer, the backbone network layer includes at least a spatial pyramid pooling feature fusion module, a centralized comprehensive convolution module and a first multi-scale feature adaptive fusion module, the neck fusion layer includes at least a path aggregation module and a second multi-scale feature adaptive fusion module; the training of the teacher network based on the labeled remote sensing data set includes:

[0022] Inputting the labeled remote sensing data set into the teacher network, performing feature extraction on the labeled remote sensing data set through the backbone network layer, wherein after pooling the image of the labeled remote sensing data set through the spatial pyramid pooling feature fusion module, convolving the pooled image through the centralized comprehensive convolution module, and adaptively fusing the convolved image through the first multi-scale feature adaptive fusion module to obtain feature maps of multiple scales;

[0023] Inputting the feature maps of the multiple scales into the neck fusion layer, performing feature fusion on the feature maps of the multiple scales through the neck fusion layer, wherein after performing multi-scale feature fusion on the feature maps of the multiple scales through the path aggregation module, the fused feature maps of the multiple scales are adaptively fused through the second multi-scale feature adaptive fusion module to obtain a feature-fused image;

[0024] The feature-fused image is input to the head detection layer, and the feature-fused image is predicted by the head detection layer.

[0025] In a possible implementation, classifying the pseudo-label dataset based on the pseudo-label generation network includes:

[0026] The confidence of the candidate box is obtained, and based on the confidence of the candidate box, a first threshold and a second threshold are set using the pseudo-label generation network, and the pseudo-label data set is classified based on the first threshold and the second threshold to obtain credible pseudo-labels, erroneous pseudo-labels and uncertain pseudo-labels.

[0027] In a possible implementation, the training of the MFPA-YOLO semi-supervised model based on the mangrove species dataset to obtain a fully trained MFPA-YOLO semi-supervised model further includes:

[0028] The total loss function of the student network is optimized to enhance the classification accuracy of the image, wherein the total loss function includes a supervised loss function and a semi-supervised loss function.

[0029] In a possible implementation, the ecological environment index includes a soil salinity adjustment factor, a tidal flat water level adjustment factor, and a soil moisture adjustment factor; the calculation formula for the mangrove carbon storage prediction value is:

[0030] ,

[0031] ,

[0032] ,

[0033] ,

[0034] in, is the predicted value of mangrove carbon storage, is the preliminary estimated value of mangrove carbon storage, is the soil salinity adjustment factor, is the tidal flat water level adjustment factor, is the soil moisture adjustment factor, is the soil salinity of the current measured sample, is the tidal flat water level of the current measured sample plot, is the soil moisture index of the current measured sample, , They are the optimum salinity and water level of mangrove species ID, is the salinity variation range parameter, is the tidal flat water level variation range parameter, , is the regression coefficient of the moisture adjustment factor.

[0035] In another aspect, the present invention further provides an electronic device, comprising: a processor and a memory;

[0036] The memory stores a computer-readable program executable by the processor;

[0037] When the processor executes the computer-readable program, the steps in the mangrove carbon storage prediction method as described above are implemented.

[0038] On the other hand, the present invention also provides a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the mangrove carbon stock prediction method as described above.

[0039] The beneficial effects of the present invention are as follows: remote sensing data are annotated based on measured data, a mangrove species data set is constructed by annotating the remote sensing data, and the cost is reduced; the constructed MFPA-YOLO semi-supervised model is trained based on the mangrove species data set, wherein the MFPA-YOLO semi-supervised model includes an input layer, a target detection layer and an output layer, the target detection layer includes a teacher network, a pseudo-label generation network and a student network, data enhancement processing is performed on some labeled remote sensing data sets and a large number of unlabeled remote sensing data sets through the teacher network and the student network of the target detection layer in the MFPA-YOLO semi-supervised model, the mangrove species characteristics are strengthened, and a multi-scale feature adaptive fusion module is introduced into the teacher network and the student network, the spatiotemporal sensitivity of the feature map after convolution is enhanced, the feature expression ability after the fusion of features of different scales is improved, the efficiency and accuracy of the MFPA-YOLO semi-supervised model are improved, the mangrove carbon reserve prediction value is determined based on the ecological environment index and the initial prediction value of the mangrove carbon reserve, and the accuracy of calculating the mangrove carbon reserve is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A flow chart of an embodiment of a method for predicting mangrove carbon reserves provided by the present invention;

[0041] Figure 2 A schematic diagram of the structure of the MFPA-YOLO semi-supervised model of the mangrove carbon storage prediction method provided by the present invention;

[0042] Figure 3 A flow chart of an embodiment of the MFPA-YOLO semi-supervised model training of the mangrove carbon storage prediction method provided by the present invention;

[0043] Figure 4 A flow chart of an embodiment of teacher network training in the MFPA-YOLO semi-supervised model of the mangrove carbon storage prediction method provided by the present invention;

[0044] Figure 5 A schematic structural diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0045] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0046] The present invention discloses a mangrove carbon reserve prediction method, electronic device and storage medium, which can be used in a computer. The method, device or computer-readable storage medium involved in the present invention can be integrated with the above-mentioned device or can be relatively independent.

[0047] A specific embodiment of the present invention discloses a mangrove carbon storage prediction method, which can be executed by a computer, specifically by one or more processors of the computer. Figure 1 As shown in Figure 2, the mangrove carbon storage prediction methods include:

[0048] S101, obtaining remote sensing data and measured data of mangroves, and annotating the remote sensing data based on the measured data to construct a mangrove species dataset;

[0049] S102, training the constructed MFPA-YOLO semi-supervised model based on the mangrove species dataset, wherein the MFPA-YOLO semi-supervised model includes an input layer, a target detection layer, and an output layer, and the target detection layer includes a teacher network, a pseudo-label generation network, and a student network;

[0050] S103. Predict the mangrove remote sensing image based on the fully trained MFPA-YOLO semi-supervised model to obtain the initial prediction value of mangrove carbon storage, determine the ecological and environmental indicators, and determine the mangrove carbon storage prediction value based on the ecological and environmental indicators and the initial prediction value of mangrove carbon storage.

[0051] Among them, the MFPA-YOLO semi-supervised model is built based on the YOLOv10 network. For the schematic diagram of the MFPA-YOLO semi-supervised model structure, please refer to Figure 2 ,like Figure 2 As shown in the figure, the MFPA-YOLO semi-supervised model includes an input layer, a target detection layer and an output layer. The input layer is the data preprocessing part, the target detection layer is the target detection part, and the output layer is the result output part. The target detection layer includes a teacher network MY-1, a pseudo-label generation network PLG and a student network MY-2. The structure of the teacher network is the same as that of the student network, including a backbone network layer, a Neck layer and a Head layer. Among them, an MFA module (multi-scale feature adaptive fusion module) is added after the SPPF module (spatial pyramid pooling feature fusion module) and the C3 module (centralized comprehensive convolution module) of the backbone network layer to enhance the spatiotemporal sensitivity of the feature map after convolution. An MFA module (multi-scale feature adaptive fusion module) is inserted after the PANet module (path aggregation module) of the Neck layer (neck fusion layer) to improve the feature expression ability after the fusion of features of different scales.

[0052] Compared with the prior art, the mangrove carbon stock prediction method provided in this embodiment obtains remote sensing data and measured data of mangroves, and annotates the remote sensing data based on the measured data to construct a mangrove species data set; reduces the cost, and trains the constructed MFPA-YOLO semi-supervised model based on the mangrove species data set, wherein the MFPA-YOLO semi-supervised model includes an input layer, a target detection layer and an output layer, and the target detection layer includes a teacher network, a pseudo-label generation network and a student network; predicts the mangrove remote sensing image to be predicted based on the well-trained MFPA-YOLO semi-supervised model, obtains the initial prediction value of mangrove carbon stock, improves the efficiency and accuracy of the MFPA-YOLO semi-supervised model, determines the ecological environment indicators, and determines the mangrove carbon stock prediction value based on the ecological environment indicators and the initial prediction value of mangrove carbon stock, thereby improving the accuracy of calculating the mangrove carbon stock.

[0053] In some embodiments, in step S101, remote sensing data and measured data of mangroves are obtained, and multi-scale high-resolution remote sensing data of mangroves are obtained within the regional scope by using unmanned aerial vehicle remote sensing equipment. Based on the regional scale, measured data of ground label plots of different mangrove distribution areas in the target area are obtained, and the measured data include mangrove species name, breast diameter, tree height and sample plot geographic coordinates. The remote sensing data is annotated based on the measured data to construct a mangrove species data set. The specific steps are as follows: first, the total amount of above-ground organisms of mangroves is calculated based on the measured data, and the initial total carbon storage of mangroves is determined based on the total amount of above-ground organisms of mangroves, that is, based on the mangrove species name, breast diameter and tree height, the total amount of above-ground organisms of mangroves of each sample plot is calculated, and the calculation formula of the total amount of above-ground organisms of mangroves is:

[0054] ,

[0055] in, is the total aboveground biomass of mangroves, is the name of the mangrove species. The first The diameter of a tree at breast height, The first The height of the trees, , , is the species-specific coefficient;

[0056] The calculation formula for the initial total carbon storage of mangroves is:

[0057] ,

[0058] in, is the initial total carbon stock of mangroves;

[0059] Secondly, the remote sensing data were labeled based on the mangrove species names, sample plot geographical coordinates and initial total carbon stock of mangroves to construct a mangrove species dataset. That is, according to the collected mangrove species names, specific geographical locations of sample plots and initial total carbon stock of mangroves, the corresponding species names, mangrove areas and carbon stocks were labeled on the remote sensing images to construct a mangrove species dataset. The mangrove species dataset includes labeled remote sensing datasets and unlabeled remote sensing datasets.

[0060] In some embodiments, in step S102, the constructed MFPA-YOLO semi-supervised model is trained based on the mangrove species dataset, and the MFPA-YOLO semi-supervised model is constructed based on the YOLOv10 network, wherein the MFPA-YOLO semi-supervised model includes three parts: an input layer, a target detection layer, and an output layer. The target detection layer includes a teacher network, a pseudo-label generation network, and a student network. The structure of the teacher network is consistent with that of the student network. The teacher network includes a backbone network layer, a Neck layer, and a Head layer. The backbone network layer includes at least an SPPF module, a C3 module, and a first MFA module, that is, a multi-scale feature adaptive fusion module MFA is added after the SPPF module and the C3 module of the YOLOv10 backbone network to enhance the spatiotemporal sensitivity of the feature map after convolution, and the Neck layer includes at least a PANet module and a second MFA module, that is, based on the Neck layer of YOLOv10, an MFA module is inserted after downsampling of each layer of PANet in the Neck layer to improve the feature expression capability after fusion of features of different scales.

[0061] In some embodiments, see Figure 3 , the MFPA-YOLO semi-supervised model is trained, the steps include:

[0062] S301, inputting the mangrove species dataset into the MFPA-YOLO semi-supervised model, training a teacher network based on the labeled remote sensing dataset, and updating the parameters of the teacher network to obtain a fully trained teacher network;

[0063] S302, predicting the unlabeled remote sensing data set based on the well-trained teacher network to obtain a pseudo-labeled data set;

[0064] S303, after classifying the pseudo-label data set based on the pseudo-label generation network, iteratively training the student network based on the classified pseudo-label data set and the unlabeled remote sensing data set, and updating the parameters of the student network based on the parameters of the teacher network to obtain a fully trained MFPA-YOLO semi-supervised model.

[0065] In some embodiments, in step S301, the mangrove species dataset is input into the MFPA-YOLO semi-supervised model, that is, the mangrove species dataset is input into the input layer, and multiple data enhancement processes are performed on part of the labeled data and a large amount of unlabeled data of the mangrove species dataset through the input layer to strengthen the characteristics of the mangrove species. Then, a teacher network is trained based on the labeled remote sensing dataset, and the parameters of the teacher network are updated to obtain a fully trained teacher network, wherein the teacher network includes a backbone network layer, a neck fusion layer (Neck layer) and a head detection layer (Head layer). The backbone network layer includes at least a spatial pyramid pooling feature fusion module (SPPF module), a centralized comprehensive convolution module (C3 module) and a first multi-scale feature adaptive fusion module (first MFA module). The neck fusion layer includes at least a path aggregation module (PANet module) and a second multi-scale feature adaptive fusion module (second MFA module). For training the teacher network, please refer to Figure 4 , the steps include:

[0066] S401, inputting the labeled remote sensing data set into the teacher network, performing feature extraction on the labeled remote sensing data set through the backbone network layer, wherein, after pooling the image of the labeled remote sensing data set through the spatial pyramid pooling feature fusion module, convolving the pooled image through the centralized comprehensive convolution module, and adaptively fusing the convolved image through the first multi-scale feature adaptive fusion module to obtain feature maps of multiple scales;

[0067] S402, inputting feature maps of multiple scales into a neck fusion layer, and performing feature fusion on the feature maps of multiple scales through the neck fusion layer, wherein after performing multi-scale feature fusion on the feature maps of multiple scales through a path aggregation module, the fused feature maps of multiple scales are adaptively fused through a second multi-scale feature adaptive fusion module to obtain a feature-fused image;

[0068] In S403, the image after feature fusion is input to the head detection layer, and the image after feature fusion is predicted by the head detection layer.

[0069] In some embodiments, in step S401, a labeled remote sensing dataset is input into the teacher network, and feature extraction is performed on the labeled remote sensing dataset through the backbone network layer, wherein the image of the labeled remote sensing dataset is pooled through the SPPF module, the pooled image is convolved through the C3 module, and the convolved image is adaptively fused through the first MFA module to obtain feature maps of multiple scales. In the SPPF module, the input feature map is subjected to one convolution operation and then three serial pooling operations are performed to fuse features of the same feature map at different scales together to enrich the semantic features of the feature map; in the C3 module, the pooled image is convolved multiple times, and in the MFA module, feature extraction is performed through two branches of convolution feature and attention feature, and then adaptive fusion is performed.

[0070] In some embodiments, in step S402, feature maps of multiple scales are input into the Neck layer, and feature fusion is performed on the feature maps of multiple scales through the Neck layer, wherein after multi-scale feature fusion is performed on the feature maps of multiple scales through the PANet module, the fused feature maps of multiple scales are adaptively fused through the second MFA module to obtain a feature fused image, and in the PANet module, information is transmitted upward layer by layer through an upsampling operation, and at each level, low-resolution feature maps are laterally connected and point-by-point added to high-resolution feature maps to integrate information of different scales, and information is transmitted downward layer by layer through a downsampling operation, and at each level, high-resolution feature maps are laterally connected and point-by-point added to low-resolution feature maps to better capture and integrate multi-scale information.

[0071] The first MFA module and the second MFA module are composed of two branches: convolutional features and attention features. In the convolutional feature branch, convolution and channel transformation are used to extract local features to obtain the convolutional feature branch output, and the calculation formula is:

[0072] ,

[0073] in, Input features for the image, is the convolution feature branch output, is the convolution function; in the attention feature branch, the channel and spatial dual-dimensional attention module is used to enhance the long-range feature dependency, and its calculation formula is:

[0074] ,

[0075] ,

[0076] ,

[0077] ,

[0078] ,

[0079] ,

[0080] in, , are the average pooling output and the maximum pooling output in the channel dimension, respectively. is the average pooling function, is the maximum pooling function, is the channel attention output, is the sigmoid activation function, It is a multi-layer perceptron, consisting of multiple fully connected layers. , are the average pooling output and the maximum pooling output of the spatial dimension respectively, is the spatial attention output; the channel attention output is weightedly fused with the spatial attention output to obtain the attention feature branch output, which is calculated as follows:

[0081] ,

[0082] in, is the output of the attention feature branch; finally, the convolution feature branch output is fused with the attention feature branch output to obtain the final feature output result, which is calculated as follows:

[0083] ,

[0084] in, is the final feature output result, which is the result of multi-scale adaptive fusion of the MFA module.

[0085] In some embodiments, in step S403, the feature-fused image is input into the Head layer, the feature-fused image is predicted through the Head layer, and multiple predictions are generated for each object through the Head layer, providing rich supervisory signals, thereby improving learning accuracy.

[0086] In some embodiments, in step S302, the unlabeled remote sensing dataset is input into a well-trained teacher network, and the unlabeled remote sensing dataset is predicted by the well-trained teacher network to obtain a pseudo-labeled dataset.

[0087] In some embodiments, in step S303, after the pseudo-label data set is classified based on the pseudo-label generation network, the student network is iteratively trained based on the classified pseudo-label data set and the unlabeled remote sensing data set, and the parameters of the student network are updated based on the parameters of the teacher network to obtain a fully trained MFPA-YOLO semi-supervised model; the specific steps of classifying the pseudo-label data set based on the pseudo-label generation network are: obtaining the candidate box confidence, based on the candidate box confidence, using the pseudo-label generation network to set the first threshold and the second threshold, classifying the pseudo-label data set based on the first threshold and the second threshold, and obtaining credible pseudo labels, erroneous pseudo labels, and Unsure pseudo-label, that is, using a high threshold (first threshold) and a low threshold (second threshold) to divide the pseudo-label dataset into credible pseudo-labels, erroneous pseudo-labels and uncertain pseudo-labels. During the training process, the teacher network continuously adjusts the network parameters to calculate the candidate box confidence. When the candidate box confidence is greater than the high threshold, the pseudo-label dataset is marked as a credible pseudo-label. When the candidate box confidence is less than the low threshold, it is marked as an erroneous pseudo-label. When the candidate box confidence is greater than or equal to the low threshold and less than or equal to the high threshold, it is marked as an uncertain pseudo-label. The high threshold and the low threshold are dynamically adjusted according to the number distribution of labeled data and unlabeled data and the number of iterations. The calculation formula is:

[0088] ,

[0089] ,

[0090] in, For the The high threshold of the number of round iterations, For the The lower threshold of the number of rounds of iterations, For the The number of rounds of iterations The pseudo-label score of the class, For the The number of round iterations counted The number of annotation labels for the class, is a parameter used to measure the boundary between unreliable pseudo labels and reliable labels. is the number of unlabeled data, is the number of labeled data.

[0091] In the MFPA-YOLO semi-supervised model training process, in order to enhance the image classification accuracy, the total loss function of the student network is optimized. The total loss function includes the supervised loss function and the semi-supervised loss function. The total loss function is composed of the loss of a series of labeled images and the loss of a single unlabeled image. The calculation formula is:

[0092] ,

[0093] in, is the total loss function, is the loss function calculated on the labeled image, that is, the supervised loss function, is the loss function calculated on the unlabeled image, that is, the semi-supervised loss function, To balance the parameters of the supervised loss function and the semi-supervised loss function;

[0094] Both the semi-supervised loss function and the supervised loss function are composed of classification loss cls, regression loss reg and confidence loss conf. The calculation formula of the supervised loss function is:

[0095] ,

[0096] in, is the supervision loss function, is the classification loss, is the regression loss, is the confidence loss, is the position in the feature map, , , is the output of the student network, , , is the pseudo label result generated, is the cross entropy loss function, is the complete cross ratio loss function; the calculation formula of its semi-supervised loss function is:

[0097] ,

[0098] ,

[0099] ,

[0100] ,

[0101] in, is the semi-supervised loss function, is the classification loss function, is the regression loss function, is the confidence loss function, , , They are the positions in the feature map respectively. The classification score, regression score and confidence score of the pseudo-label results are: is an indicator function, which outputs 1 when the condition is met, otherwise it outputs 0. for The confidence score of the candidate box is for The confidence score of the pseudo label at is the high threshold, is the low threshold.

[0102] The MFPA-YOLO semi-supervised model is trained. During the training, the parameters of the teacher network are continuously updated. After the EMA smoothing calculation is performed on the parameters of the teacher network, the parameters of the student network are updated. The calculation formula is:

[0103] ,

[0104] in, For students in the network The parameter value of the step, For the The parameter values ​​of the step teacher network, For students in the network The parameter value of the step, is the smoothing factor, which ranges from 0 to 1.

[0105] In some embodiments, in step S103, the remote sensing image of the mangrove to be predicted is predicted based on the well-trained MFPA-YOLO semi-supervised model to obtain an initial prediction value of the mangrove carbon storage, the remote sensing image of the mangrove to be predicted is input into the well-trained MFPA-YOLO semi-supervised model, the mangrove carbon storage is predicted by the well-trained MFPA-YOLO semi-supervised model, the preliminary prediction value of the mangrove carbon storage is obtained, the ecological environment index is determined, the mangrove carbon storage prediction value is determined based on the ecological environment index and the initial prediction value of the mangrove carbon storage, the influence of the ecological environment index on the true value of the carbon storage is considered, and the mangrove carbon storage prediction value of each sample plot is calculated in combination with the preliminary prediction value of the mangrove carbon storage, and the calculation formula of the mangrove carbon storage prediction value is:

[0106] ,

[0107] ,

[0108] ,

[0109] ,

[0110] in, is the predicted value of mangrove carbon storage, is the preliminary estimated value of mangrove carbon storage, is the soil salinity adjustment factor, is the tidal flat water level adjustment factor, is the soil moisture adjustment factor, is the soil salinity of the current measured sample, is the tidal flat water level of the current measured sample plot, is the soil moisture index of the current measured sample, , They are the optimum salinity and water level of mangrove species ID, is the salinity variation range parameter, is the tidal flat water level variation range parameter, , is the regression coefficient of the moisture adjustment factor.

[0111] like Figure 5 As shown, the present invention also provides an electronic device 500 , which can be a computing device such as a mobile terminal, a desktop computer, a notebook, a palmtop computer, a server, etc. The electronic device 500 includes a processor 501 , a memory 502 , and a display 503 . Figure 5 Only some components of the electronic device 500 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.

[0112] In some embodiments, the memory 502 may be an internal storage unit of the electronic device 500, such as a hard disk or memory of the electronic device 500. In other embodiments, the memory 502 may also be an external storage device of the electronic device 500, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 500. Further, the memory 502 may also include both an internal storage unit of the electronic device 500 and an external storage device. The memory 502 is used to store application software and various data installed in the electronic device 500, such as program codes installed in the electronic device 500. The memory 502 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a mangrove carbon stock prediction program is stored in the memory 502, and the mangrove carbon stock prediction program can be executed by the processor 501, thereby realizing the mangrove carbon stock prediction method of each embodiment of the present invention.

[0113] In some embodiments, the processor 501 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 502, such as a mangrove carbon storage prediction method.

[0114] In some embodiments, the display 503 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 503 is used to display identification information of the mangrove carbon storage prediction program and to display a visual user interface. The components 501-503 of the electronic device 500 communicate with each other via a system bus.

[0115] In some embodiments, when the processor 501 executes the mangrove carbon stock prediction program in the memory 502, the various steps in the mangrove carbon stock prediction method described in the above embodiments are implemented. Since the mangrove carbon stock prediction method has been described in detail above, it will not be repeated here.

[0116] Accordingly, the present invention also provides a computer-readable storage medium, which is used to store computer-readable programs or instructions. When the program or instructions are executed by a processor, the steps or functions in the mangrove carbon reserve prediction method provided by the above-mentioned method embodiments can be implemented.

[0117] In summary, the mangrove carbon stock prediction method, electronic device and storage medium provided by the present invention obtain remote sensing data and measured data of mangroves, annotate the remote sensing data based on the measured data to construct a mangrove species data set; train the constructed MFPA-YOLO semi-supervised model based on the mangrove species data set, wherein the MFPA-YOLO semi-supervised model includes an input layer, a target detection layer and an output layer, the target detection layer includes a teacher network, a pseudo-label generation network and a student network, and predict the mangrove remote sensing image to be predicted based on the well-trained MFPA-YOLO semi-supervised model to obtain an initial predicted value of mangrove carbon stock, determine ecological environmental indicators, and determine the predicted value of mangrove carbon stock based on the ecological environmental indicators and the initial predicted value of mangrove carbon stock, thereby improving the efficiency and accuracy of the model.

[0118] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0119] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting mangrove carbon reserves, characterized in that: include: Acquiring remote sensing data and measured data of mangroves, and annotating the remote sensing data based on the measured data to construct a mangrove species dataset, wherein the acquiring remote sensing data and measured data of mangroves, and annotating the remote sensing data based on the measured data to construct a mangrove species dataset includes: Calculate the total aboveground biomass of mangroves based on the measured data, and determine the initial total carbon storage of mangroves based on the total aboveground biomass of mangroves, wherein the measured data include mangrove species name, breast diameter, tree height and geographic coordinates of the sample plot; The remote sensing data are labeled based on the mangrove species name, the sample plot geographic coordinates, and the initial total carbon stock of the mangrove to construct a mangrove species dataset, wherein the mangrove species dataset includes a labeled remote sensing dataset and an unlabeled remote sensing dataset; The constructed MFPA-YOLO semi-supervised model is trained based on the mangrove species dataset, wherein the MFPA-YOLO semi-supervised model includes an input layer, a target detection layer and an output layer, the target detection layer includes a teacher network, a pseudo-label generation network and a student network, and the constructed MFPA-YOLO semi-supervised model is trained based on the mangrove species dataset, including: Inputting the mangrove species dataset into the MFPA-YOLO semi-supervised model, training the teacher network based on the labeled remote sensing dataset, and updating the parameters of the teacher network to obtain a fully trained teacher network; Predicting the unlabeled remote sensing dataset based on the fully trained teacher network to obtain a pseudo-labeled dataset; After classifying the pseudo-label data set based on the pseudo-label generation network, iteratively training the student network based on the classified pseudo-label data set and the unlabeled remote sensing data set, and updating the parameters of the student network based on the parameters of the teacher network to obtain a fully trained MFPA-YOLO semi-supervised model; Based on the well-trained MFPA-YOLO semi-supervised model, the remote sensing images of mangroves to be predicted are predicted to obtain the initial predicted value of mangrove carbon storage, determine the ecological environmental indicators, and determine the predicted value of mangrove carbon storage based on the ecological environmental indicators and the initial predicted value of mangrove carbon storage.

2. The method for predicting mangrove carbon reserves according to claim 1, characterized in that: The calculation formula for the total amount of above-ground organisms in mangroves is: ， in, is the total aboveground biomass of mangroves, is the name of the mangrove species. The first The diameter of a tree at breast height, The first The height of the trees, , , is the coefficient of species; The calculation formula for the initial total carbon storage of mangroves is: ， in, is the initial total carbon storage of mangroves.

3. The method for predicting mangrove carbon reserves according to claim 1, characterized in that: The teacher network has the same structure as the student network. The teacher network includes a backbone network layer, a neck fusion layer and a head detection layer. The backbone network layer includes at least a spatial pyramid pooling feature fusion module, a centralized comprehensive convolution module and a first multi-scale feature adaptive fusion module. The neck fusion layer includes at least a path aggregation module and a second multi-scale feature adaptive fusion module. The step of training the teacher network based on the labeled remote sensing data set includes: Inputting the labeled remote sensing data set into the teacher network, performing feature extraction on the labeled remote sensing data set through the backbone network layer, wherein after pooling the image of the labeled remote sensing data set through the spatial pyramid pooling feature fusion module, convolving the pooled image through the centralized comprehensive convolution module, and adaptively fusing the convolved image through the first multi-scale feature adaptive fusion module to obtain feature maps of multiple scales; Inputting the feature maps of the multiple scales into the neck fusion layer, performing feature fusion on the feature maps of the multiple scales through the neck fusion layer, wherein after performing multi-scale feature fusion on the feature maps of the multiple scales through the path aggregation module, the fused feature maps of the multiple scales are adaptively fused through the second multi-scale feature adaptive fusion module to obtain a feature-fused image; The feature-fused image is input to the head detection layer, and the feature-fused image is predicted by the head detection layer.

4. The method for predicting mangrove carbon reserves according to claim 3, characterized in that: The classifying the pseudo-label data set based on the pseudo-label generation network includes: The confidence of the candidate box is obtained, and based on the confidence of the candidate box, a first threshold and a second threshold are set using the pseudo-label generation network, and the pseudo-label data set is classified based on the first threshold and the second threshold to obtain credible pseudo-labels, erroneous pseudo-labels and uncertain pseudo-labels.

5. The method for predicting mangrove carbon reserves according to claim 3, characterized in that: The method of training the MFPA-YOLO semi-supervised model based on the mangrove species dataset to obtain a fully trained MFPA-YOLO semi-supervised model further includes: The total loss function of the student network is optimized to enhance the classification accuracy of the image, wherein the total loss function includes a supervised loss function and a semi-supervised loss function.

6. The method for predicting mangrove carbon reserves according to claim 1, characterized in that: The ecological environment indicators include soil salinity adjustment factor, tidal flat water level adjustment factor and soil moisture adjustment factor; the calculation formula for the mangrove carbon storage prediction value is: ， ， ， ， in, is the predicted value of mangrove carbon storage, is the preliminary estimated value of mangrove carbon storage, is the soil salinity adjustment factor, is the tidal flat water level adjustment factor, is the soil moisture adjustment factor, is the soil salinity of the current measured sample, is the tidal flat water level of the current measured sample plot, is the soil moisture index of the current measured sample, , They are the optimum salinity and water level of mangrove species ID, is the salinity variation range parameter, is the tidal flat water level variation range parameter, , is the regression coefficient of the moisture adjustment factor.

7. An electronic device, characterized in that: including memory and processor; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the mangrove carbon storage prediction method according to any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the mangrove carbon stock prediction method as described in any one of claims 1-6.

Images