A method for estimating forest canopy carbon assimilation based on water-carbon coupling theory

Abstract

This invention belongs to the field of ecology and environmental science and provides a forest canopy carbon assimilation estimation method based on water-carbon coupling theory. The method includes: collecting forest remote sensing images and radar point cloud data, segmenting individual tree canopies, extracting forest tree data, generating a monitoring equipment layout map, acquiring sensor data, calculating raw canopy carbon assimilation rate estimates, and generating forest canopy carbon assimilation estimation results. By utilizing an equipment layout generation model to generate the monitoring equipment layout map, the method optimizes the spatial arrangement of monitoring equipment and improves the reliability of collected data. Furthermore, through the forest canopy carbon assimilation estimation optimization model, the method takes into account the impact of forest remote sensing images, avoids the errors introduced by conventional linear fitting methods, and improves the generalizability of the method.

Description

Technical Field

[0001] The present invention relates to the fields of ecology and environmental science, and in particular to a forest canopy carbon assimilation estimation method based on water-carbon coupling theory. Background Art

[0002] Forests are key ecosystems regulating global climate, and their carbon cycle processes directly influence atmospheric CO2 concentrations and climate change trends. Through photosynthesis, forests absorb CO2 and convert it into organic matter for storage, a mechanism crucial for mitigating the greenhouse effect. The forest canopy is the primary interface for gas exchange with the outside world, and its CO2 and water vapor fluxes determine a forest's carbon sequestration capacity and water use efficiency. Accurately estimating these fluxes is crucial for understanding the carbon cycle and predicting ecological responses.

[0003] However, traditional methods have problems such as high cost and complex operation, making it difficult to meet long-term monitoring needs. In addition, traditional tree monitoring requires manual measurement, which is time-consuming, labor-intensive and inefficient. At the same time, since only dominant tree species are monitored, the influence of a few tree species is ignored, which will lead to large errors between the estimated results and the actual values. Since the forest is in a dynamic state, the use of conventional linear fitting will result in poor fitting accuracy. Summary of the Invention

[0004] In order to overcome the shortcomings of the existing technology, the purpose of the present invention is to provide a forest canopy carbon assimilation estimation method based on water-carbon coupling theory to solve the problems of large errors in conventional linear fitting methods and unreasonable layout of monitoring equipment.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for estimating forest canopy carbon assimilation based on water-carbon coupling theory includes:

[0007] Use drones equipped with cameras and LiDAR radars to collect remote sensing data of the target monitoring forest, obtaining forest remote sensing images and radar point cloud data;

[0008] performing single tree crown segmentation on the forest remote sensing image to obtain a single tree segmentation image;

[0009] Extracting crown information of each single tree sub-image in the single tree segmentation image, and extracting height information corresponding to each single tree sub-image based on the radar point cloud data to obtain forest tree data;

[0010] Inputting the forest tree data into a pre-trained equipment layout plan generation model to obtain monitoring equipment layout data, and mapping the monitoring equipment layout data to the forest remote sensing image to obtain a monitoring equipment layout map;

[0011] Deploying monitoring sensors in the target monitoring forest according to the monitoring equipment layout diagram, and collecting data using the monitoring sensors to obtain sensor collected data;

[0012] Calculating the sensor collected data using a pre-built canopy carbon assimilation rate estimation formula to obtain raw canopy carbon assimilation rate estimation data;

[0013] The sensor collected data, the original canopy carbon assimilation rate estimation data and the forest remote sensing image are input into a forest canopy carbon assimilation estimation optimization model for data processing to obtain a forest canopy carbon assimilation estimation result.

[0014] Preferably, performing single tree crown segmentation on the forest remote sensing image to obtain a single tree segmentation image includes:

[0015] Converting the radar point cloud data into a canopy height model, and segmenting the canopy height model using a watershed segmentation algorithm to obtain an initial canopy outline candidate area;

[0016] performing adaptive histogram equalization on the forest remote sensing image to obtain a canopy enhanced image;

[0017] Performing pixel-level registration on the initial tree crown outline candidate area and the canopy enhanced image through affine transformation according to the position and attitude information of the UAV to obtain a registered image;

[0018] The U-Net network is used to extract, fuse and decode the spectral features, structural features and boundary-aware features of the registered image to obtain a single tree crown probability map and instance mask;

[0019] counting the number of masks and the average probability of the pixel to be analyzed within a preset neighborhood range based on the single tree crown probability map and the instance mask, and determining the area where the pixel to be analyzed is located as a candidate area when the number of masks and the average probability are greater than a preset standard;

[0020] Calculating the volume centroid of each candidate area according to the canopy height model, and constructing a triangulated network using a Delaunay triangulation algorithm based on the volume centroid to obtain a Delaunay triangulated network;

[0021] Determining the candidate areas corresponding to the two endpoints of each edge in the Delaunay triangulation as a group of overlapping tree crown pairs, and determining overlapping areas of the overlapping tree crown pairs;

[0022] extracting a cutting line of the overlapping area using a height gradient;

[0023] Calculating the point cloud density of each pixel within a preset range of the cutting line, and if the difference in the point cloud density of adjacent pixel points is greater than a density difference threshold, determining the adjacent pixel points as boundary points;

[0024] Calculating the distances between the boundary points and the two volume centroids respectively, and classifying the boundary points into the candidate region corresponding to the volume centroid with the smallest distance, to obtain an image segmentation result;

[0025] A preset tree species priori knowledge base is used to match geometric feature constraints on each block in the image segmentation result to obtain the single tree segmentation image determined by the tree species type.

[0026] Preferably, the forest tree data includes: tree ID, tree species type, crown area, crown width, tree height, crown base height, canopy volume, spectral characteristics and location coordinates.

[0027] Preferably, the training process of the equipment layout plan generation model includes:

[0028] Collecting forest tree data of a target research forest area, and constructing a virtual forest model using a three-dimensional modeling tool based on the forest tree data;

[0029] Performing parameter simulation on the forest tree data of the virtual forest model by adjusting environmental variables to obtain forest simulation data;

[0030] Inputting the forest simulation data into a preset graph neural network to generate a layout plan and obtain intermediate device layout data;

[0031] Evaluate the intermediate equipment deployment data using a preset deployment plan evaluation formula to obtain a plan evaluation value;

[0032] The scheme evaluation value is used to optimize the parameters of the graph neural network to obtain the iteratively completed equipment layout scheme generation model.

[0033] Preferably, the canopy carbon assimilation rate estimation formula is expressed as:

[0034] ;

[0035] in, is the estimated result of canopy carbon assimilation rate; is the transpiration rate per unit leaf area; is the water density; is the water vapor gas constant; Represents temperature data; is the forest leaf area index; It is a water vapor pressure deficit; is the ambient carbon dioxide concentration; Stable carbon isotope resolution for plant leaves.

[0036] Preferably, the training process of the forest canopy carbon assimilation estimation optimization model includes:

[0037] Arrange the monitoring sensors in the target area according to the monitoring equipment layout diagram, and use the monitoring sensors to obtain sensor data collected by the target area;

[0038] Utilizing the UAV to collect multispectral images of the target area according to a fixed altitude and route planning;

[0039] Using eddy covariance method to obtain the overall canopy carbon assimilation data of the target area;

[0040] Collecting and calculating the sensor data according to the monitoring equipment layout diagram to obtain the original canopy carbon assimilation rate estimation data;

[0041] Performing geometric correction and radiometric calibration on the multispectral image, and performing outlier removal and time stamp alignment on the sensor collected data;

[0042] Pre-training a ResNet18 network using the ImageNet image set, and extracting features from the multispectral image using the convolutional layer of the ResNet18 network to obtain an image feature vector;

[0043] performing standardization processing on the sensor collected data to obtain standard sensor data;

[0044] performing splicing processing on the image feature vector, the standard sensor data, and the original canopy carbon assimilation rate estimation data to obtain a fused feature vector;

[0045] Inputting the fused feature vector into a multi-layer fully connected network for prediction to obtain overall canopy carbon assimilation prediction data;

[0046] The mean square error of the overall canopy carbon assimilation data and the overall canopy carbon assimilation prediction data is calculated to obtain a loss value, and the network parameters of the multi-layer fully connected network are iterated using the loss value to obtain the iteratively completed forest canopy carbon assimilation estimation optimization model.

[0047] Preferably, the monitoring sensor comprises: a thermal diffusion probe measuring device.

[0048] Preferably, the U-Net network is used to extract, fuse and decode the spectral features, structural features and boundary-aware features of the registered image to obtain a single tree crown probability map and an instance mask, including:

[0049] Extracting texture features, color features, and spectral depth features from the multispectral channel data of the registered image using the convolutional layer of the U-Net network to obtain the spectral features;

[0050] Using the pooling layer of the U-Net network to reduce the spatial resolution of the registered image to obtain the structural features;

[0051] Using the Sobel operator to identify the edge of the tree crown in the registered image to obtain the boundary perception feature;

[0052] Performing jump connections on the spectral features, the structural features, and the boundary perception features to obtain cross-layer fusion features;

[0053] Upsampling the cross-layer fusion features and concatenating feature maps to obtain fused multi-scale information;

[0054] The decoder uses a Sigmoid activation function to predict the crown probability of the fused multi-scale information to obtain the single tree crown probability map, and uses a Softmax activation function to assign instance labels to the fused multi-scale information to obtain the instance mask.

[0055] Preferably, the tree species prior knowledge base includes: crown morphology data, proportion data and structure data.

[0056] Preferably, the environmental variables include: ambient temperature, precipitation pattern, solar radiation, atmospheric carbon dioxide concentration, altitude, slope, aspect, soil moisture content, nutrient concentration, wind speed and canopy competition intensity.

[0057] The present invention discloses the following technical effects:

[0058] The present invention provides a forest canopy carbon assimilation estimation method based on water-carbon coupling theory. By using an equipment layout plan generation model to generate a monitoring equipment layout map, the problem of unreasonable monitoring equipment layout is solved and the spatial layout of the monitoring equipment is optimized. Through the forest canopy carbon assimilation estimation optimization model, the problem of large errors in conventional linear fitting methods is solved, and the original canopy carbon assimilation rate estimation data is optimized using sensor collected data and forest remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0060] Figure 1 A schematic diagram of a process for estimating forest canopy carbon assimilation based on water-carbon coupling theory provided in an embodiment of the present invention;

[0061] Figure 2 A schematic diagram of a process for obtaining a single tree segmentation image provided by an embodiment of the present invention;

[0062] Figure 3 A schematic diagram of the equipment deployment solution generation model training process provided in an embodiment of the present invention;

[0063] Figure 4 A schematic diagram of the training process of the forest canopy carbon assimilation estimation optimization model provided by an embodiment of the present invention;

[0064] Figure 5 A schematic diagram of the process of obtaining a single tree crown probability map and instance mask provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] The purpose of the present invention is to provide a forest canopy carbon assimilation estimation method based on water-carbon coupling theory to solve the problems of large errors in conventional linear fitting methods and unreasonable layout of monitoring equipment.

[0067] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0068] Figure 1 A schematic diagram of the forest canopy carbon assimilation estimation process based on the water-carbon coupling theory provided in an embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides a method for estimating forest canopy carbon assimilation based on water-carbon coupling theory, comprising:

[0069] Step 100: Using a drone equipped with a camera and a LiDAR radar to collect remote sensing data of the target monitoring forest, obtaining a forest remote sensing image and radar point cloud data;

[0070] Step 200: performing single tree crown segmentation on the forest remote sensing image to obtain a single tree segmentation image;

[0071] Step 300: extracting crown information of each single tree sub-image in the single tree segmentation image, and extracting height information corresponding to each single tree sub-image based on the radar point cloud data to obtain forest tree data;

[0072] Step 400: Inputting the forest tree data into a pre-trained equipment layout plan generation model to obtain monitoring equipment layout data, and mapping the monitoring equipment layout data to the forest remote sensing image to obtain a monitoring equipment layout map;

[0073] Step 500: Deploying monitoring sensors in the target monitoring forest according to the monitoring equipment layout diagram, and utilizing the monitoring sensors to collect data to obtain sensor collected data;

[0074] Step 600: Calculate the sensor collected data using a pre-built canopy carbon assimilation rate estimation formula to obtain raw canopy carbon assimilation rate estimation data;

[0075] Step 700: Input the sensor collected data, the original canopy carbon assimilation rate estimation data, and the forest remote sensing image into a forest canopy carbon assimilation estimation optimization model for data processing to obtain a forest canopy carbon assimilation estimation result.

[0076] refer to Figure 2 , performing single tree crown segmentation on the forest remote sensing image to obtain a single tree segmentation image, including:

[0077] Step 201: converting the radar point cloud data into a canopy height model, and segmenting the canopy height model using a watershed segmentation algorithm to obtain an initial tree crown outline candidate area;

[0078] Step 202: performing adaptive histogram equalization on the forest remote sensing image to obtain a canopy enhanced image;

[0079] Step 203: performing pixel-level registration of the initial tree crown outline candidate area and the canopy enhanced image through affine transformation according to the position and attitude information of the UAV to obtain a registered image;

[0080] Step 204: Using a U-Net network, extract, fuse, and decode the spectral features, structural features, and boundary-aware features of the registered image to obtain a single tree crown probability map and an instance mask.

[0081] Step 205: Counting the number of masks and the average probability of the pixel to be analyzed within a preset neighborhood range based on the single tree crown probability map and the instance mask; when the number of masks and the average probability are greater than a preset standard, determining the area where the pixel to be analyzed is located as a candidate area;

[0082] Step 206: Calculate the volume centroid of each candidate area according to the canopy height model, and construct a triangulated network using the Delaunay triangulation algorithm based on the volume centroid to obtain a Delaunay triangulated network;

[0083] Step 207: determining the candidate regions corresponding to the two endpoints of each edge in the Delaunay triangulation as a group of overlapping tree crown pairs, and determining the overlapping regions of the overlapping tree crown pairs;

[0084] Step 208: extracting the cutting line of the overlapping area using the height gradient;

[0085] Step 209: Calculate the point cloud density of each pixel within the preset range of the cutting line. If the difference in the point cloud density of adjacent pixels is greater than a density difference threshold, determine the adjacent pixels as boundary points.

[0086] Step 210: Calculate the distance between the boundary point and the two volume centroids respectively, and assign the boundary point to the candidate region corresponding to the volume centroid with the smallest distance, to obtain an image segmentation result;

[0087] Step 211: Use a preset tree species prior knowledge base to match geometric feature constraints on each block in the image segmentation result to obtain the single tree segmentation image determined by the tree species type.

[0088] Preferably, the forest tree data includes: tree ID, tree species type, crown area, crown width, tree height, crown base height, canopy volume, spectral characteristics and location coordinates.

[0089] refer to Figure 3 The training process of the equipment layout plan generation model includes:

[0090] Step 401: collecting forest tree data of a target research forest area, and constructing a virtual forest model using a three-dimensional modeling tool based on the forest tree data;

[0091] Step 402: performing parameter simulation on the forest tree data of the virtual forest model by adjusting environmental variables to obtain forest simulation data;

[0092] Step 403: Input the forest simulation data into a preset graph neural network to generate a layout plan and obtain intermediate device layout data;

[0093] Step 404: Evaluate the intermediate device deployment data using a preset deployment plan evaluation formula to obtain a plan evaluation value;

[0094] Step 405: Optimize the parameters of the graph neural network using the solution evaluation value to obtain the iteratively completed equipment layout solution generation model.

[0095] Furthermore, the canopy carbon assimilation rate estimation formula is expressed as follows:

[0096] ;

[0097] in, is the estimated result of canopy carbon assimilation rate; is the transpiration rate per unit leaf area; is the water density; is the water vapor gas constant; Represents temperature data; is the forest leaf area index; It is a water vapor pressure deficit; is the ambient carbon dioxide concentration; Stable carbon isotope resolution for plant leaves.

[0098] refer to Figure 4 The training process of the forest canopy carbon assimilation estimation optimization model includes:

[0099] Step 701: Arrange the monitoring sensors in the target area according to the monitoring equipment layout diagram, and use the monitoring sensors to obtain sensor data collected in the target area;

[0100] Step 702: using the UAV to collect multispectral images of the target area according to a fixed altitude and route planning;

[0101] Step 703: using eddy covariance method to obtain the overall canopy carbon assimilation data of the target area;

[0102] Step 704: collecting and calculating the sensor data according to the monitoring equipment layout diagram to obtain the original canopy carbon assimilation rate estimation data;

[0103] Step 705: performing geometric correction and radiometric calibration on the multispectral image, and performing outlier removal and time stamp alignment on the sensor collected data;

[0104] Step 706: Pre-training a ResNet18 network using the ImageNet image set, and extracting features from the multispectral image using the convolutional layer of the ResNet18 network to obtain an image feature vector;

[0105] Step 707: performing standardization processing on the sensor collected data to obtain standard sensor data;

[0106] Step 708: performing splicing processing on the image feature vector, the standard sensor data, and the original canopy carbon assimilation rate estimation data to obtain a fused feature vector;

[0107] Step 709: Input the fused feature vector into a multi-layer fully connected network for prediction to obtain overall canopy carbon assimilation prediction data;

[0108] Step 710: Calculate the mean square error between the overall canopy carbon assimilation data and the overall canopy carbon assimilation prediction data to obtain a loss value, and use the loss value to iterate the network parameters of the multi-layer fully connected network to obtain the iterated forest canopy carbon assimilation estimation optimization model.

[0109] Preferably, the monitoring sensor comprises: a thermal diffusion probe measuring device.

[0110] refer to Figure 5 , using the U-Net network to extract, fuse and decode the spectral features, structural features and boundary-aware features of the registered image to obtain a single tree crown probability map and instance mask, including:

[0111] Step 20401: Using the convolutional layer of the U-Net network, extracting texture features, color features, and spectral depth features from the multispectral channel data of the registered image to obtain the spectral features;

[0112] Step 20402: Using the pooling layer of the U-Net network to reduce the spatial resolution of the registered image to obtain the structural features;

[0113] Step 20403: using the Sobel operator to identify the tree crown edge of the registered image to obtain the boundary perception feature;

[0114] Step 20404: performing jump connections on the spectral features, the structural features, and the boundary perception features to obtain cross-layer fusion features;

[0115] Step 20405: upsampling the cross-layer fusion features and concatenating feature maps to obtain fused multi-scale information;

[0116] Step 20406: Using a Sigmoid activation function, the decoder performs crown probability prediction on the fused multi-scale information to obtain the single tree crown probability map, and using a Softmax activation function, the decoder performs instance label assignment on the fused multi-scale information to obtain the instance mask.

[0117] Optionally, the tree species prior knowledge base includes: crown morphology data, proportion data and structure data.

[0118] Preferably, the environmental variables include: ambient temperature, precipitation pattern, solar radiation, atmospheric carbon dioxide concentration, altitude, slope, aspect, soil moisture content, nutrient concentration, wind speed and canopy competition intensity.

[0119] Specifically, a thermal diffusion probe measurement system was installed to observe trunk sap flow and obtain trunk sap flow data. Real-time meteorological data of the forest area was obtained, such as atmospheric temperature (T), rainfall (P), relative humidity (RH), total radiation (TBQ), photosynthetically active radiation (PAR), and saturation vapor pressure deficit (VPD). VPD was obtained using the following empirical formula:

[0120]

[0121]

[0122] Where, Represents the saturated water vapor pressure difference at temperature T; a, b, and c are all empirical parameters, a is 0.611, b is 17.502, and c is 240.97.

[0123] Furthermore, the trunk sap flow density is calculated based on the Granier model and empirical formula. The trunk sap flow density can be expressed as:

[0124]

[0125] Where Js is the liquid flow density, is the instantaneous temperature difference between the upper and lower probes, The maximum temperature difference within a day;

[0126] Preferably, based on the sap flow density, the method of converting whole tree transpiration to canopy transpiration, the whole tree transpiration ( ) is calculated as follows:

[0127]

[0128] Where As is the sapwood area of ​​trees in this diameter class.

[0129] Canopy transpiration of a particular tree species ( )for:

[0130]

[0131] Where, is the average sap flow density of trees of diameter class i in the stand, is the sum of the sapwood areas of trees of type i.

[0132] The transpiration rate per unit leaf area of ​​the forest stand ( ):

[0133]

[0134] Where LAI is the forest leaf area index, For the total area.

[0135] Furthermore, the canopy stomatal conductance was calculated based on Kostner's simplified formula ( ) can be expressed as:

[0136]

[0137] Where Gv is the gas constant for water vapor, ρ is the density of water, and D is the water vapor pressure deficit.

[0138] Preferably, leaf samples at different canopy heights and air samples near the canopy are collected during the plant growing season and non-growing season to determine the stable carbon isotope composition of the leaves and air. , The value is determined based on PDB (PeeDee Bele-mnite) as the standard and can be expressed as:

[0139]

[0140] Where: Indicates sample The percentage of deviation from the standard sample in parts per thousand; Indicates the carbon ratio of the standard substance PDB.

[0141] The calculation method for the stable carbon isotope resolution Δ of plant leaves is:

[0142]

[0143] Where: is the carbon isotope ratio of atmospheric carbon dioxide.

[0144] Furthermore, the water-carbon coupling model at the leaf scale was constructed based on Farquhar's stable carbon isotope model for C3 plants, which can be expressed as follows:

[0145]

[0146]

[0147]

[0148]

[0149] Where, The canopy carbon dioxide uptake rate, is the stomatal conductance of CO2, is the stomatal conductance of water, a is the diffusion process of gas under the pore 13 C / 12 The discrimination rate of C is poor, and b is in the carboxylation process 13 C / 12 The discrimination rate of C is poor. represents the intercellular carbon dioxide concentration, represents the ambient carbon dioxide concentration.

[0150] Specifically, a canopy-scale water-carbon coupling model was constructed. By using tree trunk sap flow monitoring, continuous forest canopy transpiration can be obtained, and continuous canopy average stomatal conductance can be calculated. The canopy carbon assimilation rate can be estimated using the following formula:

[0151]

[0152]

[0153] Specifically, the drone platform should be capable of carrying multiple payloads. The payload interface must be compatible with simultaneous operation of the camera and LiDAR radar, supporting timestamp synchronization. The camera should be a multispectral camera or a high-resolution visible light camera. The LiDAR radar should be a medium-range 3D laser radar with a scanning frequency of no less than 200kHz and a vertical field of view consistent with the camera.

[0154] Furthermore, the flight area was delineated by using online maps to determine the target forest's boundaries, terrain, and vegetation distribution. The flight altitude was set based on sensor performance and tree canopy height. Typically, the effective LiDAR scanning altitude is 50-150 meters (to avoid signal attenuation). A fixed altitude was maintained during flight to avoid altitude differences. A zigzag flight path was planned, with an overlap ratio of at least 80% in the heading direction and at least 70% in the lateral direction. The flight speed was approximately 8 meters per second.

[0155] Ideally, collect multispectral imagery (including RGB, red-edge, and near-infrared bands) of the target area at a fixed altitude and route plan, ensuring that a single acquisition covers the entire study area (e.g., 1 km x 1 km). Record the acquisition timestamp (accurate to the second) and the drone's position and attitude data for subsequent geometric correction. Evenly distribute 5-10 monitoring points across the study area to collect tree trunk sap flow rate, record temperature, relative humidity, carbon dioxide concentration, light intensity, and measure leaf area index in real time.

[0156] Furthermore, based on POS data and ground control points, image registration is performed through ENVI / ArcGIS to eliminate the distortion caused by the UAV's flight posture; pixel values ​​are converted into actual reflectivity to generate radiometrically corrected remote sensing images.

[0157] Optimally, the network outputs individual tree crown probability maps and instance masks. These data represent pixel-level segmentation results, essentially providing a preliminary assessment of the crown area within the image. Subsequent processing is based on these preliminary segmentation results. The neural network output provides the foundational data for subsequent post-processing, which further optimizes and refines these data.

[0158] Furthermore, the neural network outputs individual tree crown probability maps and instance masks to identify overlapping crowns. The probability maps and instance masks are analyzed to identify areas of potential crown overlap. These areas typically exhibit blurred boundaries between adjacent masks or overlapping probability values ​​within a certain range. Based on the three-dimensional spatial distribution of the LiDAR point cloud, the canopy volume centroid of each candidate region (i.e., a possible individual tree crown region) is calculated. A Delaunay triangulation is then constructed to describe the spatial relationships between trees, providing a geometric basis for subsequent separation of overlapping crowns.

[0159] Specifically, let the probability map of a single tree crown be P(x,y), where (x,y) represents the image pixel coordinates, and the value range is 0≤P(x,y)≤1. The closer the value is to 1, the higher the probability that the pixel belongs to the crown; the instance mask is M(x,y), and different integers represent different crown instances. In order to determine the overlapping area, we first define a probability threshold And a 3×3 neighborhood N(x,y) centered at (x,y). For each pixel (x,y), calculate the number of different instance masks n(x,y) in its neighborhood and the average probability When n(x,y)>1 and > , the area where the pixel is located is considered to be an overlapping area.

[0160] Furthermore, the graph structure of the proximity relationship is modeled. A triangulated mesh constructed from a set of centroids transforms the spatial relationship between trees into non-intersecting triangle connections. The vertices (centroids) of any two adjacent triangles represent spatially adjacent tree crowns. The tree crown pairs (i, j) corresponding to the triangulated mesh edges are potential overlapping pairs. Subsequent separation operations process the intersection of these adjacent tree crowns.

[0161] Optimally, equipment placement plans should consider the correlation between tree spatial location, species characteristics, and canopy structure. Graph neural networks (GNNs) can model forest tree data as a graph structure, effectively capturing the spatial dependencies and structural characteristics between trees. This makes them suitable for handling equipment placement problems with spatial correlation. Node features include: tree ID, species type, crown area, crown width, height, canopy volume, and location coordinates; edge features include: Euclidean distance between two trees, canopy overlap, and species correlation (ecological correlations derived from a priori knowledge base); the graph structure of GNNs uses an undirected graph structure.

[0162] Specifically, this embodiment provides a deployment scheme evaluation formula for the network model training process, which is expressed as follows:

[0163]

[0164]

[0165]

[0166]

[0167]

[0168] in, 、 、 、 They are uniformity weight, representativeness weight, diversity weight and interference weight; 、 、 、 They are spatial uniformity index, tree species representativeness index, environmental heterogeneity diversity index, and equipment interference index; is the coefficient of variation of the area; 、 are the standard deviation and mean area of ​​the region, respectively; 、 、 are the tree species weight, the number of individuals of the kth tree species, and the total number of individuals of the kth tree species respectively; is the total number of tree species; is the proportion of devices in the mth environmental gradient interval; is the total number of environmental gradient intervals; The number of device pairs whose distance is less than the minimum reasonable spacing; The total number of devices.

[0169] The beneficial effects of the present invention are as follows:

[0170] The present invention generates a monitoring equipment layout diagram by utilizing an equipment layout plan generation model, thereby optimizing the spatial arrangement of monitoring equipment and improving the reliability of collected data. Through the forest canopy carbon assimilation estimation optimization model, the impact of forest remote sensing images is taken into account, thus avoiding the errors caused by conventional linear fitting methods and improving the generalization of the method.

[0171] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0172] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for estimating forest canopy carbon assimilation based on water-carbon coupling theory, characterized in that: include: Use drones equipped with cameras and LiDAR radars to collect remote sensing data of the target monitoring forest, obtaining forest remote sensing images and radar point cloud data; performing single tree crown segmentation on the forest remote sensing image to obtain a single tree segmentation image; Extracting crown information of each single tree sub-image in the single tree segmentation image, and extracting height information corresponding to each single tree sub-image based on the radar point cloud data to obtain forest tree data; Inputting the forest tree data into a pre-trained equipment layout plan generation model to obtain monitoring equipment layout data, and mapping the monitoring equipment layout data to the forest remote sensing image to obtain a monitoring equipment layout map; Deploying monitoring sensors in the target monitoring forest according to the monitoring equipment layout diagram, and collecting data using the monitoring sensors to obtain sensor collected data; Calculating the sensor collected data using a pre-built canopy carbon assimilation rate estimation formula to obtain raw canopy carbon assimilation rate estimation data; Inputting the sensor collected data, the original canopy carbon assimilation rate estimation data, and the forest remote sensing image into a forest canopy carbon assimilation estimation optimization model for data processing to obtain a forest canopy carbon assimilation estimation result; The training process of the equipment layout plan generation model includes: Collecting forest tree data of a target research forest area, and constructing a virtual forest model using a three-dimensional modeling tool based on the forest tree data; Performing parameter simulation on the forest tree data of the virtual forest model by adjusting environmental variables to obtain forest simulation data; Inputting the forest simulation data into a preset graph neural network to generate a layout plan and obtain intermediate device layout data; Evaluate the intermediate equipment deployment data using a preset deployment plan evaluation formula to obtain a plan evaluation value; The scheme evaluation value is used to optimize the parameters of the graph neural network to obtain the iteratively completed equipment layout scheme generation model.

2. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 1, characterized in that: Performing single tree crown segmentation on the forest remote sensing image to obtain a single tree segmentation image includes: Converting the radar point cloud data into a canopy height model, and segmenting the canopy height model using a watershed segmentation algorithm to obtain an initial canopy outline candidate area; performing adaptive histogram equalization on the forest remote sensing image to obtain a canopy enhanced image; Performing pixel-level registration on the initial tree crown outline candidate area and the canopy enhanced image through affine transformation according to the position and attitude information of the UAV to obtain a registered image; The U-Net network is used to extract, fuse and decode the spectral features, structural features and boundary-aware features of the registered image to obtain a single tree crown probability map and instance mask; counting the number of masks and the average probability of the pixel to be analyzed within a preset neighborhood range based on the single tree crown probability map and the instance mask, and determining the area where the pixel to be analyzed is located as a candidate area when the number of masks and the average probability are greater than a preset standard; Calculating the volume centroid of each candidate area according to the canopy height model, and constructing a triangulated network using a Delaunay triangulation algorithm based on the volume centroid to obtain a Delaunay triangulated network; Determining the candidate areas corresponding to the two endpoints of each edge in the Delaunay triangulation as a group of overlapping tree crown pairs, and determining overlapping areas of the overlapping tree crown pairs; extracting a cutting line of the overlapping area using a height gradient; Calculating the point cloud density of each pixel within a preset range of the cutting line, and if the difference in the point cloud density of adjacent pixel points is greater than a density difference threshold, determining the adjacent pixel points as boundary points; Calculating the distances between the boundary points and the two volume centroids respectively, and classifying the boundary points into the candidate region corresponding to the volume centroid with the smallest distance, to obtain an image segmentation result; A preset tree species priori knowledge base is used to match geometric feature constraints on each block in the image segmentation result to obtain the single tree segmentation image determined by the tree species type.

3. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 1, characterized in that: The forest tree data includes: tree ID, tree species type, crown area, crown width, tree height, crown base height, canopy volume, spectral characteristics and location coordinates.

4. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 1, characterized in that: The canopy carbon assimilation rate estimation formula is expressed as follows: ； in, is the estimated result of canopy carbon assimilation rate; is the transpiration rate per unit leaf area; is the water density; is the water vapor gas constant; Represents temperature data; is the forest leaf area index; It is a water vapor pressure deficit; is the ambient carbon dioxide concentration; Stable carbon isotope resolution for plant leaves.

5. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 1, characterized in that: The training process of the forest canopy carbon assimilation estimation optimization model includes: Arrange the monitoring sensors in the target area according to the monitoring equipment layout diagram, and use the monitoring sensors to obtain sensor data collected by the target area; Utilizing the UAV to collect multispectral images of the target area according to a fixed altitude and route planning; Using eddy covariance method to obtain the overall canopy carbon assimilation data of the target area; Collecting and calculating the sensor data according to the monitoring equipment layout diagram to obtain the original canopy carbon assimilation rate estimation data; Performing geometric correction and radiometric calibration on the multispectral image, and performing outlier removal and time stamp alignment on the sensor collected data; Pre-training a ResNet18 network using the ImageNet image set, and extracting features from the multispectral image using the convolutional layer of the ResNet18 network to obtain an image feature vector; performing standardization processing on the sensor collected data to obtain standard sensor data; performing splicing processing on the image feature vector, the standard sensor data, and the original canopy carbon assimilation rate estimation data to obtain a fused feature vector; Inputting the fused feature vector into a multi-layer fully connected network for prediction to obtain overall canopy carbon assimilation prediction data; The mean square error of the overall canopy carbon assimilation data and the overall canopy carbon assimilation prediction data is calculated to obtain a loss value, and the network parameters of the multi-layer fully connected network are iterated using the loss value to obtain the iteratively completed forest canopy carbon assimilation estimation optimization model.

6. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 1, characterized in that: The monitoring sensor includes: a thermal diffusion probe measuring device.

7. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 2, characterized in that: The U-Net network is used to extract, fuse and decode the spectral features, structural features and boundary-aware features of the registered image to obtain a single tree crown probability map and instance mask, including: Extracting texture features, color features, and spectral depth features from the multispectral channel data of the registered image using the convolutional layer of the U-Net network to obtain the spectral features; Using the pooling layer of the U-Net network to reduce the spatial resolution of the registered image to obtain the structural features; Using the Sobel operator to identify the edge of the tree crown in the registered image to obtain the boundary perception feature; Performing jump connections on the spectral features, the structural features, and the boundary perception features to obtain cross-layer fusion features; Upsampling the cross-layer fusion features and concatenating feature maps to obtain fused multi-scale information; The decoder uses a Sigmoid activation function to predict the crown probability of the fused multi-scale information to obtain the single tree crown probability map, and uses a Softmax activation function to assign instance labels to the fused multi-scale information to obtain the instance mask.

8. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 2, characterized in that: The tree species prior knowledge base includes: crown morphological data, proportion data and structure data.

9. The method for estimating forest canopy carbon assimilation based on water-carbon coupling theory according to claim 1, characterized in that: The environmental variables include: ambient temperature, precipitation patterns, solar radiation, atmospheric carbon dioxide concentration, altitude, slope, aspect, soil moisture content, nutrient concentration, wind speed, and canopy competition intensity.

Images