A forestry survey planning, design and analysis method based on 3D laser modeling

Abstract

The present invention relates to the technical field of forestry management data processing, and particularly to a forestry survey planning, design and analysis method based on three-dimensional laser modeling. The present invention includes the steps of: collecting lidar point cloud data, multispectral images and thermal infrared images of a target area through an unmanned aerial vehicle, a ground mobile platform and satellite remote sensing; establishing corresponding analysis packages based on different analysis scenarios, adaptively selecting corresponding algorithms for each analysis package to perform data compression and dimensionality reduction, and then fusing the data; for each analysis package, constructing a corresponding three-dimensional model using the fused data, and extracting the characteristic parameters required for forestry surveys from the three-dimensional model; fusing the analysis packages, analyzing and constructing the extracted characteristic parameters, and providing a basis for forestry planning, design and analysis. The present invention solves the limitation of poor adaptability in traditional forestry surveys through the technical link of "scenario-based analysis package construction - differential processing - cross-scale fusion".

Description

Technical Field

[0001] The present invention relates to the technical field of forestry management data processing, and particularly to a method for forestry survey planning, design and analysis based on three-dimensional laser modeling. Background Art

[0002] The key problem faced by current forestry resource surveys lies in the lack of flexibility and scene adaptability in data processing methods. Traditional methods use fixed processes to handle all types of survey tasks, which can neither meet the needs of large-scale rapid censuses nor achieve fine analysis of individual trees. In existing forestry resource surveys, if data integrity is retained, it will lead to low efficiency in processing massive data in macro monitoring scenarios. If data is forcibly compressed, key details will be lost in the analysis of individual tree characteristics. This black-and-white processing mode of the existing technology forces forestry workers to make compromises between efficiency and accuracy, seriously restricting the practical value of survey results. A deeper problem lies in the fragmented management and analysis of multi-source data. Existing technologies often independently process data collected by different technical means such as lidar, satellite images, and thermal infrared, lacking the ability for cross-scale correlation analysis. For example, it is difficult to establish a quantitative relationship between terrain feature data and the growth status of individual trees, and the tree health information reflected by thermal infrared images cannot effectively assist in overall forest area planning decisions. This data island phenomenon not only causes waste of information resources but also leads to one-sidedness in comprehensive analysis conclusions.

[0003] Therefore, the existing technology system sacrifices data details when pursuing processing speed excessively, while blindly retaining complete data requires a large amount of computing resources. This contradiction often places forestry survey work in a dilemma, where the reliability of quickly obtained conclusions is insufficient, and the high-precision analysis results lose their application value due to the lack of timeliness, ultimately affecting the scientific nature and timeliness of planning, design and analysis. Summary of the Invention

[0004] In view of the deficiencies of the existing technology, the present invention provides a method for forestry survey planning, design and analysis based on three-dimensional laser modeling, which solves the limitations of "single data, fixed algorithm, and fragmented scene" in traditional forestry surveys.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions:

[0006] A forestry survey planning, design and analysis method based on three-dimensional laser modeling, comprising the steps of: collecting lidar point cloud data, multispectral images and thermal infrared images of a target area through drones, ground mobile platforms and satellite remote sensing; establishing corresponding analysis packages based on different analysis scenarios, the analysis packages including an analysis package for the macro monitoring scenario of forest resources and an analysis package for the research scenario of individual tree characteristics, and adaptively selecting corresponding algorithms for each analysis package to perform data compression and dimensionality reduction, and then fusing the lidar point cloud data with the multispectral images and thermal infrared images; for each analysis package, constructing a corresponding three-dimensional model using the fused data, and extracting characteristic parameters required for forestry surveys from the three-dimensional model; fusing the two analysis packages, and analyzing and constructing the extracted characteristic parameters to provide a basis for forestry planning, design and analysis.

[0007] Further, for the analysis package of the macro monitoring scenario of forest resources, data compression is to perform joint adjustment processing on the lidar point cloud data, multispectral images and thermal infrared images to eliminate redundant data; data dimensionality reduction is to calculate the correlation between the original multi-source data and the macro characteristics of the forest, set a threshold for the correlation coefficient, retain the relevant characteristic data above the threshold, and eliminate the relevant characteristic data below the threshold.

[0008] Further, the data fusion of the analysis package for the macro monitoring scenario of forest resources includes:

[0009] Denoise the lidar point cloud and extract macro terrain features and vegetation features, convert the pixel DN value of the multispectral image into radiance value and then calculate spectral features, convert the gray value of the thermal infrared image into actual temperature value and then extract energy distribution features; during fusion, first preliminarily align the three types of data according to geographical coordinates, then perform secondary matching based on feature similarity, adopt weighted fusion, allocate weights according to the importance of data sources, and fuse the matched feature data according to the allocated weights.

[0010] Further, constructing a three-dimensional model for the analysis package of the macro monitoring scenario of forest resources includes:

[0011] Use the macro terrain features and vegetation features extracted from the lidar point cloud as the spatial skeleton, endow the model with vegetation details and classification information through the multispectral image, endow the model with energy distribution through the thermal infrared image, construct a three-dimensional model, and render the model surface using the texture and color information of the multispectral image.

[0012] Further, the extraction of characteristic parameters for the analysis package of the macro monitoring scenario of forest resources includes:

[0013] The forest coverage rate is obtained by calculating the ratio of the space volume occupied by the tree point cloud in the three-dimensional model to the space volume of the entire target area; the average tree height is calculated by obtaining the height data of the tree point cloud in the three-dimensional model and then using the weighted average method, and the terrain slope is obtained by performing plane fitting analysis on the terrain point cloud in the three-dimensional model.

[0014] Furthermore, in the analysis package for the study scenario of tree individual characteristics, data compression is to calculate the covariance matrix of the lidar point cloud data, analyze the variance ratio of each dimension, select the dimension with a large variance ratio as the core segmentation dimension, divide intervals according to the coordinate range of the core segmentation dimension, compress each interval independently, introduce distance weights to divide nodes, cluster the point cloud within each interval based on the weights, identify the core points, retain the core points and their neighborhood points, and eliminate low-density noise points; data dimensionality reduction is to sample and split various characteristics of the lidar point cloud data of a single tree, screen out the key characteristics, set a threshold, and retain the characteristics that exceed the threshold.

[0015] Furthermore, the data fusion of the analysis package for the study scenario of tree individual characteristics includes:

[0016] Denoise the lidar point cloud data, multi-spectral image, and thermal infrared image of a single tree respectively. Then, determine the trunk central axis and position by extracting the characteristics of the lidar point cloud data of a single tree, and then calculate the trunk diameter, distinguish the crown levels, and obtain the branch characteristics; extract the texture characteristics from the multi-spectral image of a single tree. When matching, identify the tree contour through the multi-spectral image, then match it with the corresponding lidar point cloud through geometric position consistency, and then associate the thermal infrared and lidar data, and perform weighted fusion on the geometric characteristics of the lidar data, the texture and health index of the multi-spectral, and the temperature characteristics of the thermal infrared.

[0017] Furthermore, constructing a three-dimensional model for the analysis package of tree individual characteristics research includes: based on the fused and denoised lidar point cloud data, multi-spectral image, and thermal infrared image data, using point cloud-based surface reconstruction, dynamically adjusting the sphere radius parameter according to the local curvature of the tree point cloud, and at the same time, performing texture mapping on the model surface by combining the texture information of the multi-spectral image.

[0018] Furthermore, the extraction of characteristic parameters for the analysis package of tree individual characteristics research includes: directly obtaining the trunk height by measuring the length of the trunk central axis in the three-dimensional model; calculating the crown shape index according to the three-dimensional shape of the crown; statistically analyzing the number of branch bifurcations through topological analysis of the branch skeleton structure in the three-dimensional model.

[0019] Further, fuse the two analysis packages and analyze and construct the extracted feature parameters. Specifically, associate the parameters of the two analysis packages through a unified geographic coordinate system, establish an index using a spatial database, configure feature weights according to scenario requirements, construct a macro-micro parameter model, generate a corresponding scenario map, and visualize it through a 3D model.

[0020] On the other hand, the present invention discloses an electronic device, comprising:

[0021] at least one processor; and

[0022] a memory communicatively connected to the at least one processor; wherein,

[0023] the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned forestry survey planning design and analysis method based on 3D laser modeling.

[0024] The present invention provides a forestry survey planning design and analysis method based on 3D laser modeling, which has the following beneficial effects:

[0025] Based on the technical route of "scenario analysis package construction - differential processing - cross-scale fusion", the present invention designs dynamic adaptation data processing strategies for two typical scenarios of forest resource macro monitoring and single-tree fine analysis: in the macro monitoring scenario, through multi-source data joint adjustment and intelligent dimensionality reduction technology, while ensuring the accuracy of key statistical features, significantly improve the processing efficiency of massive data; in the individual feature analysis scenario, adopt a vertical dimension segmentation and feature screening mechanism to completely retain details such as crown levels and branch topologies, providing a high-fidelity data basis for single-tree feature analysis.

[0026] Through the cross-scale data fusion mechanism, the present invention realizes the in-depth collaborative processing of lidar point clouds, multi-spectral images and thermal infrared data. Based on the spatial association and dynamic weight configuration of a unified geographic coordinate system, establish a quantitative mapping relationship between macro-topographic parameters and tree individual features. This global-local linkage analysis mode not only supports scientific decision-making for overall forest area planning, but also enables accurate positioning and health diagnosis of abnormal single trees, significantly improving the reliability of resource assessment and disaster warning.

[0027] In addition, the present invention balances data accuracy and computational load through elastic architecture design, breaking through the passive choice of "efficiency for accuracy" or "accuracy for efficiency" in traditional methods. The 3D laser modeling technology combined with multi-source data fusion optimizes the entire process from forest area terrain reconstruction to single-tree morphology analysis, providing adaptive decision support for dynamic forest resource monitoring and promoting the transformation of forestry management from experience-driven to data-intelligence-driven. Description of the Drawings

[0028] Figure 1 This is a flowchart of the forestry survey, planning, design and analysis method based on 3D laser modeling of the present invention;

[0029] Figure 2 This is the visualization diagram of the 3D model in the embodiment of the present invention. Specific Embodiments

[0030] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0031] A forestry survey, planning, design and analysis method based on 3D laser modeling, as Figure 1 shown, includes the steps of: collecting lidar point cloud data, multispectral images and thermal infrared images of the target area through drones, ground mobile platforms and satellite remote sensing; in implementation, a drone is equipped with a GPS device, a lidar and an optical camera. During flight, the GPS device continuously collects the coordinates of ground control points, together with the lidar point cloud data and multispectral images of the target area, and at the same time transmits the lidar point cloud data and multispectral images to the on-board data processing system in real time. The on-board data processing system monitors the collected data in real time; the ground mobile platform collects thermal infrared images and optical images of the target area according to the driving route; for data collection in complex terrains and no-fly zones for drones, the latest satellite remote sensing data is obtained for supplementation. Specifically, the drone data collection is based on a preset forest area boundary (GIS vector data) to generate a terrain-following flight route, and the flight height (5-10 m from the forest canopy) is dynamically adjusted according to the terrain undulation to ensure that the lidar point cloud density is ≥ 20 points / m². The lidar and the multispectral camera are synchronously collected through hardware triggering. The GPS records the POS data every 0.1 second and aligns it with the time stamp of the point cloud data, and the lidar point cloud and the multispectral image are transmitted back to the on-board processing system in real time. The ground platform is equipped with a FLIRT865 thermal imager and travels along the forest road, and the thermal infrared image and the optical image are synchronously collected. The dynamic supplementation of satellite remote sensing data is achieved by accessing Sentinel 2 and WorldView-3 satellite data, automatically downloading the latest images according to the coordinates of the drone no-fly zone, and in steep slope (slope > 45°) areas, fusing the DSM data generated by the satellite stereo pair with the drone point cloud, and filling the terrain blind area of the drone through Poisson surface reconstruction.

[0032] Based on different analysis scenarios, corresponding analysis packages are established. The analysis packages include the analysis package for the macro monitoring scenario of forest resources and the analysis package for the research scenario of individual tree characteristics. Each analysis package adaptively selects the corresponding algorithm for data compression and dimensionality reduction. In implementation, for the analysis package of the macro monitoring scenario of forest resources, data compression is to perform joint adjustment processing on lidar point cloud data, multi-spectral images, and thermal infrared images. Specifically, for the analysis package of the macro monitoring scenario of forest resources, first, align the lidar point cloud (including GPS / IMU positioning data), multi-spectral image (including POS data), and thermal infrared image (including timestamp) through the UTC time axis; use the GPS ground control point coordinates of the unmanned aerial vehicle (UAV) airborne system as the unified spatial reference, and map the three types of data to the WGS84 coordinate system. The lidar point cloud and the multi-spectral image are matched by SIFT feature points (extract at least 500 pairs of homologous points), and the affine transformation matrix is calculated (error < 1 pixel); the thermal infrared image is upsampled to the resolution of the multi-spectral image through bilinear interpolation, and then preliminarily aligned based on the geographical coordinate grid (10m × 10m).

[0033] In data compression, first construct a joint adjustment optimization model, including:

[0034] Define the joint adjustment objective function as: where L i , S i , T i are the observed values of the i-th feature point of the lidar point cloud, multi-spectral image, and thermal infrared image respectively;

[0035] , , are the optimized theoretical values;

[0036] α, β, γ are weight assignments.

[0037] Then, use the Levenberg-Marquardt algorithm to iteratively solve, optimize the spatial coordinates of the lidar point cloud, the radiance value of the multi-spectral image, and the temperature value of the thermal infrared image until the objective function converges (residual reduction rate < 1e-5).

[0038] Then, grid the registered lidar point cloud, retain the median elevation points in each grid, and eliminate redundant points; for the multi-spectral image, eliminate redundant bands through band correlation analysis. For example, when the correlation between the blue and green bands > 0.9, only retain the green band; perform regression analysis on the thermal infrared temperature data and the multi-spectral NDVI value, and eliminate the data dimensions of the regions with insignificant temperature changes (slope < 0.1); perform joint clustering on the lidar point cloud intensity value and the multi-spectral near-infrared band, and merge the point cloud clusters with similar radiation characteristics.

[0039] In the embodiment, a large forest area is divided into 1 km × 1 km sub-blocks, each block is independently subjected to joint adjustment processing, a 10 m overlap band is set at the edge of the block, and the splicing gaps are eliminated through Kriging interpolation.

[0040] For the analysis package of forest resource macro-monitoring scenarios, data dimensionality reduction is to calculate the correlation between the original multi-source data and the macro characteristics of the forest, set the threshold of the correlation coefficient, retain the relevant feature data above the threshold, and eliminate the relevant feature data below the threshold. Specifically, data dimensionality reduction is to first normalize the laser radar point cloud intensity value to [0,1], convert the multispectral image radiation brightness value into reflectance based on the radiation calibration parameter, standardize the thermal infrared temperature value to Z-score, use the processed data as input data, assign weights to the input data, and spatially grid the weighted data. The grid can be set to 10m×10m. The statistical features in each grid include the laser radar elevation median, point cloud density, intensity variance features, multispectral NDVI mean, red edge band reflectance features, thermal infrared temperature mean, day and night temperature difference features; then perform macro feature label matching, and associate each grid with a preset macro feature label (such as forest coverage, terrain slope). Specifically, calculate the correlation coefficient between each input feature (such as NDVI mean, point cloud density) and the target macro feature (such as coverage, slope). The calculation formula is: r represents the correlation coefficient, which indicates the degree of linear correlation between two variables x and y. The value range of r is −1≤r≤1, where r=1 indicates a perfect positive correlation, r=−1 indicates a perfect negative correlation, and r=0 indicates no linear correlation. i ,y i Represents the observation value of the i-th sample in the dataset. x is the input feature, including the median elevation value, point cloud density, intensity variance characteristics of the lidar, the NDVI mean value and red edge band reflectivity characteristics of the multispectral, the temperature mean value and day and night temperature difference characteristics of the thermal infrared. i represents the i-th sample observation value of variable X (input feature). For example, in the embodiment, it can correspond to the i-th data of a specific feature such as the median elevation of a lidar, point cloud density, or the mean NDVI of a multispectral image. y is the target macro feature, including coverage and slope. Then y i Represents the i-th sample observation value of variable Y (target macro-feature). For example, it corresponds to the i-th data of macro-features such as forest coverage and terrain slope. represents the sample mean of variable X, that is, all x i The average value of represents the sample mean of variable y, that is, all y i The average value of .

[0041] Set a threshold (|r|≥0.3), retain strongly correlated features (such as NDVI and coverage rate r = 0.65, point cloud density and slope r = 0.41), and eliminate weakly correlated features (such as the diurnal temperature difference of thermal infrared and coverage rate r = 0.12).

[0042] For the analysis package in the macro monitoring scenario of forest resources, after data compression and dimensionality reduction, the lidar point cloud data is fused with multi-spectral images and thermal infrared images. For the lidar point cloud, an outlier removal algorithm based on statistical analysis is used. By calculating the average distance from a point to its K nearest neighbors, with the mean plus 3 times the standard deviation as the threshold, the noise points generated by measurement errors are removed, and the macro terrain features and vegetation features are extracted. For the multi-spectral image, according to the camera radiometric calibration parameters, the pixel DN value is converted into radiance value, and the spectral features are calculated. For the thermal infrared image, the gray value is converted into the actual temperature value, and the energy distribution features are extracted. During fusion, the three types of data are initially aligned according to the geographical coordinates, and then secondarily matched based on feature similarity, and weighted fusion is adopted, with weights assigned according to the importance of the data sources.

[0043] During implementation, for the noise points in the lidar point cloud, calculate the average distance from a point to its K nearest neighbors, adopt the K-nearest neighbor outlier detection, set K = 50, and calculate the average distance d from each point to its nearest 50 neighbors avg ; statistically calculate the global average distance μ d and the standard deviation σ d , set the threshold T = μ d + 3σ d ; eliminate all noise points with d avg > T (such as bird and dust reflection points).

[0044] For the radiometric correction of the multi-spectral image, according to the camera radiometric calibration parameters (gain g, offset b), the pixel DN value is converted into radiance L, and the calculation formula is: L = g ⋅ DN + b; then the spectral features are extracted, including NDVI, SAVI, and REIP.

[0045] For the thermal infrared temperature conversion, based on the sensor calibration parameters, the gray value Gray is converted into the actual temperature T, and the calculation formula is: T = k ⋅ Gray + c (k, c are calibration coefficients); then the energy distribution features are extracted, including the temperature mean, variance, and diurnal temperature difference (which requires combining data collected multiple times).

[0046] When performing data fusion, the lidar point cloud, multi-spectral image, and thermal infrared data are unified to the WGS84 coordinate system with a grid resolution of 10m × 10m. Using the UAV GPS ground control point coordinates (accuracy of ±2cm) as the reference, the position deviation of the satellite data is corrected.

[0047] Secondary feature matching includes LiDAR and multispectral alignment, and thermal infrared and LiDAR alignment. Among them, the steps of LiDAR and multispectral alignment are as follows: within the grid, extract SIFT feature points (≥20 pairs per grid), calculate the affine transformation matrix, and the registration error < 1 pixel. Perform bilinear interpolation on the multispectral image to match the LiDAR point cloud resolution.

[0048] The steps of thermal infrared and LiDAR alignment are as follows: based on the temperature-elevation correlation, perform non-rigid registration through maximizing mutual information. In steep slope areas (slope > 25°), use the ICP algorithm (Iterative Closest Point) to optimize the matching accuracy.

[0049] The weight assignment strategy for weighted fusion is as follows: LiDAR has the highest spatial accuracy and dominates the topographic and vegetation structure features, so the LiDAR weight is 0.5. The multispectral provides spectral classification information but has a lower resolution, so the multispectral weight is 0.3. The thermal infrared reflects the energy distribution but is vulnerable to weather interference, so the thermal infrared weight is 0.2.

[0050] Feature-level fusion generates a fused feature vector for each grid (10m × 10m). The feature vector includes the median elevation, slope, mean NDVI, mean temperature, point cloud density, red-edge reflectance, and is calculated by weighted summation according to the weights:

[0051] Feature value = 0.5 ⋅ LiDAR feature + 0.3 ⋅ multispectral feature + 0.2 ⋅ thermal infrared feature.

[0052] During implementation, dynamic weight adjustment can be performed. For example, if the quality of thermal infrared data is poor (such as cloud cover > 50%), its weight is reduced to 0.1, and the LiDAR weight is increased to 0.6.

[0053] Construct a 3D model for the forest resources macro monitoring and analysis package, including:

[0054] Extract the macro topographic and vegetation features from the LiDAR point cloud to serve as the spatial framework. Endow the model with vegetation details and classification information through the multispectral image, extract the energy distribution through the thermal infrared image, construct the 3D model using the Poisson surface reconstruction algorithm, and render the model surface using the texture and color information of the multispectral image. Obtain the forest coverage rate by calculating the ratio of the spatial volume occupied by the tree point cloud in the 3D model to the spatial volume of the entire target area. Obtain the average tree height by obtaining the height data of the tree point cloud in the 3D model and then calculating by the weighted average method. Obtain the terrain slope by performing plane fitting analysis on the terrain point cloud in the 3D model.

[0055] In the extraction of macroscopic terrain features from lidar points, the elevation model, slope, and aspect are calculated by the moving surface fitting method. In the specific implementation, the cloth simulation filtering (CSF) algorithm is used to separate ground points from non-ground points (vegetation, buildings). The cloth stiffness coefficient k = 0.5 and the number of iterations n = 100 are set, and the point cloud is divided into a ground layer and a non-ground layer. For steep slope areas (slope > 25°), k = 0.3 is adjusted to enhance the retention of terrain details. The ground point cloud is divided into 10m × 10m grids, and the 20% points with the lowest elevation in each grid are retained as reference ground points. The fitting window size is adaptively adjusted according to the terrain complexity. The window size in flat areas (elevation variance < 1m) is 5 × 5 grids (50m × 50m); the window size in complex terrains (elevation variance ≥ 1m) is 3 × 3 grids (30m × 30m).

[0056] For the ground points in each window, the terrain is fitted with a quadratic surface model, and the calculation model is:

[0057] z = ax 2 + by 2 + cxy + dx + ey + f;

[0058] In the formula, a represents the quadratic curvature in the x direction, reflecting the bending degree of the terrain on the x-axis; b represents the quadratic curvature in the y direction, reflecting the bending degree of the terrain on the y-axis; c represents the cross curvature in the xy direction, describing the twisting degree of the terrain in the x-y plane; d represents the linear slope in the x direction, that is, the inclination rate of the terrain in the x-axis direction; e represents the linear slope in the y direction, that is, the inclination rate of the terrain in the y-axis direction; f represents the reference elevation at the coordinate origin.

[0059] The coefficients a, b, c, d, e, and f are solved by the least squares method; the residual threshold is set to 0.1m, and the points exceeding the threshold are regarded as noise and removed. The elevation value of the fitted surface is used as the elevation of the grid center point to generate a digital surface model (DSM) with a resolution of 0.5m; for the windows where the fitting fails (such as no ground points), Kriging interpolation is used to complete the data from adjacent grids, and finally a DSM with a resolution of 0.5m is generated. The 3 × 3 Sobel operator is applied to the DSM to calculate the gradient G x in the east-west direction and the gradient G y in the north-south direction;

[0060] Then the slope s is calculated, and the calculation formula for the slope s is: ;

[0061] The aspect calculation is to perform a fast Fourier transform on the DSM to extract the terrain frequency domain features; the calculation formula for the aspect A (unit: degree, 0° is due north, increasing clockwise) is: ; For high-frequency noise areas (FFT amplitude > threshold), Gaussian filtering is used for smoothing and then recalculated.

[0062] For extracting vegetation features from lidar points, based on the lidar point cloud after outlier removal, the CSF algorithm (k = 0.5, n = 100) is used again to separate ground and vegetation points. Normalize the intensity values of the vegetation points, statistically analyze the height histogram, and identify the peak intervals of the herb layer, shrub layer, and tree layer through the Gaussian mixture model (GMM). Calculate the layering parameters, calculate the median and standard deviation of the intensity for each layer of point clouds, and remove the outlier points whose intensity exceeds the median ± 2 standard deviations. For the tree layer, use DBSCAN clustering (neighborhood radius 1m, minimum number of points 10) to distinguish individual tree crowns and remove isolated points; for the shrub layer, combine the multi-spectral NDVI value (>0.4) to verify the vegetation density, and mark NDVI < 0.3 as dead shrubs; for the herb layer, exclude non-vegetation points through thermal infrared temperature (during the day > ambient temperature + 2°C). Gridify (10m × 10m) to statistically analyze the point cloud density of each layer and generate a spatial distribution map.

[0063] Use the Poisson surface reconstruction algorithm to construct a 3D model. Specifically, downsample the lidar point cloud (voxel size 0.5m) to reduce the computational load. Use the SIFT feature points of the multi-spectral image to register with the lidar point cloud (error < 1 pixel) to ensure the alignment of texture and geometric structure. Adopt the Poisson surface reconstruction algorithm, use the lidar point cloud as the input, and control the grid fineness by adjusting the depth parameter. For areas with dense vegetation, increase the number of iterations to optimize the surface smoothness. Map texture information such as NDVI and red-edge band reflectance of the multi-spectral image to the surface of the 3D model. Convert the DN value of the multi-spectral image to reflectance through radiometric calibration parameters, and then generate a pseudo-color map based on the mean value within the grid and overlay it on the model surface to enhance the visualization effect of vegetation classification.

[0064] The calculation of forest coverage rate is to extract the tree point cloud in the 3D model through threshold segmentation (such as height > 1m), calculate the occupied spatial volume, and estimate the total target area volume based on the DSM range and average terrain height.

[0065] The calculation steps of the average tree height are as follows: Extract the highest point height of each individual tree crown from the tree point cloud of the 3D model, and combine the GMM layering results to assign weights to the point clouds in different height intervals (such as weight 0.7 for the tree layer and 0.3 for the shrub layer).

[0066] The verification of terrain slope is to perform local plane fitting on the terrain point cloud, select a 3×3 grid neighborhood, fit the plane equation ax + by + cz + d = 0 through the least squares method, calculate the normal vector of the fitted plane, and then deduce the slope. For error correction, compare the DSM slope with the plane fitting slope. If the difference exceeds 5°, recheck the ground point cloud separation parameters and adjust the rigidity coefficient or the number of iterations of the CSF algorithm to ensure the accuracy of slope calculation.

[0067] In the analysis package for the research scenario of individual tree characteristics, data compression first analyzes the variances of each dimension of the tree point cloud data through the covariance matrix, selects the vertical dimension for segmentation, and at the same time introduces distance weights to divide nodes, preserves local characteristics, and evaluates the effect with the compression ratio and feature loss rate, and dynamically adjusts the construction parameters accordingly; data dimensionality reduction first samples and splits various characteristics of the tree point cloud through a random forest, screens out key features, weights the distances of the key feature dimensions, adjusts the perplexity, and verifies the dimensionality reduction results through visualization and quantitative analysis.

[0068] The data compression of the analysis package for the research scenario of individual tree characteristics includes the steps:

[0069] Calculate the covariance matrix for the LiDAR point cloud data of a single tree (including dimensions such as three-dimensional coordinates, intensity values, and echo times), and analyze the variance contributions of each dimension. Including:

[0070] First calculate the mean of each dimension, and the formula is: ;

[0071] Then calculate the elements of the covariance matrix:

[0072] The element C ij of the covariance matrix C is: ;

[0073] Finally, the covariance matrix is expressed as: ;

[0074] In the formula: Z k represents the k-th sample vector of the LiDAR point cloud data of a single tree, including dimension information such as three-dimensional coordinates, intensity values, and echo times; Z ik represents the specific value of the i-th dimension in the k-th sample. For example, when i = 1, it can represent the coordinate value, and when i = 4, it can represent the intensity value; μ i represents the mean of the i-th dimension, such as calculating the average value of the vertical coordinates in all sample point cloud data; n represents the number of samples of the LiDAR point cloud data of a single tree; m represents the total number of dimensions of the LiDAR point cloud data of a single tree (such as the total number of dimensions of three-dimensional coordinates, intensity values, echo times, etc.), and μ represents the vector composed of the means of each dimension

[0075] μ i and reflects the average characteristics of each dimension of the point cloud data.

[0076] By comparing the variances of each dimension, the vertical direction is preferentially selected as the segmentation benchmark because it contributes the most to structural features such as tree height and crown stratification. The hierarchical segmentation strategy divides the tree point cloud into intervals such as the trunk area (e.g., z ≤ 2m), the basic crown layer (2m < z ≤ 5m), and the upper crown layer (z > 5m) according to the range of the z-axis coordinates (vertical direction) of the point cloud, and each interval is compressed independently.

[0077] The steps of the distance-weighted division node are to first construct a local neighborhood, that is, for each point cloud p i , taking it as the center, search for neighborhood points within a radius r = 0.5m, and calculate the average distance d between this point and the neighborhood points avg , then introduce the distance-weight function ω i , , where σ is the scale parameter (e.g., σ = 0.2m). The greater the weight, the higher the contribution of this point to the local structure and the higher the retention priority. Then perform node division, that is, cluster the point cloud within each layer based on the weight, use the DBSCAN algorithm (neighborhood radius 0.3m, minimum number of points 5) to identify the core points, retain the core points and their neighborhood points, and eliminate the low-density noise points.

[0078] For the data dimensionality reduction of the analysis package for the research scenario of tree individual characteristics, it includes the steps:

[0079] Extract multi-dimensional features of the point cloud of a single tree, including geometric features, radiation features, and topological features. Among them, geometric features include point cloud density, local curvature, and vertical height difference. Radiation features include lidar intensity value and multi-spectral NDVI mean. Topological features include neighborhood point distribution entropy and skeleton branch number. Then use random forest training, with individual parameters such as trunk diameter and crown shape index as labels, to construct a random forest model. Calculate the importance score of each feature through Gini impurity, screen the key features, set a threshold (e.g., the top 30% of the importance scores), and retain the features strongly related to the individual parameters. For example, retain local curvature (correlation coefficient r = 0.78 with trunk diameter), vertical height difference (correlation coefficient r = 0.65 with crown shape), etc.

[0080] Through the above data compression and dimensionality reduction strategies, the analysis package for tree individual characteristics research can efficiently reduce data redundancy while retaining the morphological and physiological details of a single tree, providing lightweight and highly distinguishable feature inputs for subsequent 3D modeling and parameter extraction.

[0081] The data fusion of the analysis package for the research scenario of tree individual characteristics includes:

[0082] Denoise the lidar point cloud data, multispectral image, and thermal infrared image respectively. Then, the lidar point cloud determines the trunk center axis and position through PCA, calculates the trunk diameter using the least squares method, distinguishes the canopy layers by DBSCAN, and obtains the branch features through skeleton extraction and topological analysis; the multispectral image extracts texture features using the gray-level co-occurrence matrix, judges the tree health by NDVI. During matching, the multispectral image identifies the tree contour through edge detection and morphological processing, and then matches it with the lidar point cloud according to the trunk height feature. By establishing a temperature-diameter relationship model, the thermal infrared and lidar data are correlated.

[0083] In specific implementation, the statistical outlier removal algorithm is used to denoise the lidar point cloud. Calculate the average distance of each point to its K nearest neighbors (K = 30), and use the mean plus 3 times the standard deviation as the threshold to remove noise points (such as birds, sensor error points). For the trunk area, lower the threshold (mean + 2 times the standard deviation) to retain details. The non-local means filter is applied to denoise the multispectral image, reducing noise using the similarity of image patches. The parameter settings are a search window of 15×15 pixels, a neighborhood window of 5×5 pixels, and a gray-level difference weight of 0.8. The bilateral filter is used to denoise the thermal infrared image, considering both spatial distance and gray-level difference. The parameter settings are a spatial Gaussian kernel standard deviation of 3 and a gray-level Gaussian kernel standard deviation of 0.1.

[0084] Feature extraction of the lidar point cloud is to perform PCA analysis on the point cloud of a single tree, extract the principal components, project the point cloud onto the principal component plane (the first two principal components), and fit a line as the trunk center axis through the RANSAC algorithm. Specific steps: Calculate the covariance matrix of the point cloud, obtain the eigenvectors and eigenvalues. Select the direction with the largest eigenvalue as the trunk extension direction (Z-axis). Fit a line to the point cloud in the Z-axis direction to obtain the trunk center axis.

[0085] Trunk diameter calculation is to extract the points closest to the center axis in the vertical directions on both sides of the center axis, and fit a circle through the least squares method.

[0086] Apply the DBSCAN algorithm to the non-trunk point cloud (Z-axis height > 2m), with parameter settings of a neighborhood radius of 1.5m and a minimum number of points of 15. Distinguish different canopy layers (such as the main canopy layer, side branch layer) according to the clustering results. Use a point cloud-based skeleton extraction algorithm (such as VoxelSlicing) to generate the branch skeleton structure. Statistically analyze the number of skeleton nodes and branches through topological analysis to identify branch bifurcation points.

[0087] Feature extraction of multi - spectral images includes: calculating the gray - level co - occurrence matrix (GLCM) of multi - spectral images, extracting texture features such as contrast, entropy, and correlation. The window size is set to 11×11 pixels, and the step size is 5 pixels. Calculate NDVI through the red band and the near - infrared band. Perform edge detection on multi - spectral images (Canny operator, threshold 100 - 200), and combine morphological dilation (kernel size 3×3) and contour tracking to extract tree contours.

[0088] The association between thermal infrared and lidar is to establish a regression model of thermal infrared temperature and trunk diameter, realizing the association between thermal infrared temperature data and lidar geometric features. In specific implementation, more than 200 trees of different tree species with a breast - diameter range of 5 - 50 cm are selected as training samples in the plot, and the following data are collected synchronously:

[0089] Lidar features: trunk diameter (D, unit: cm), trunk height (H, unit: m), average point - cloud intensity (I);

[0090] Thermal infrared features: average trunk temperature (T mean , unit: °C), daily temperature peak T peak , unit: °C), day - night temperature difference ΔT, unit: °C);

[0091] Environmental parameters: atmospheric temperature (T a ), unit: °C), relative humidity (RH, unit: %), solar radiation intensity (S, unit: W / m²).

[0092] The regression model of thermal infrared temperature and trunk diameter is constructed as: ;

[0093] where α 0 - α 5 are regression coefficients, ϵ is the error term. After training and optimizing the above - mentioned model, use the data of 50 trees not involved in training to verify the model, calculate the absolute error between the measured diameter and the predicted diameter, and control the absolute error within the mean < 2 cm and the standard deviation < 1.5 cm, meeting the accuracy requirements of forestry surveys.

[0094] In multi - source data matching, for the matching of trunk height, it is based on the trunk height extracted by lidar, search for the contour area corresponding to the height in the multi - spectral image, and perform matching through geometric position consistency. Feature - level fusion is to perform weighted fusion of lidar geometric features (such as trunk diameter, crown volume), multi - spectral texture and health index, and thermal - infrared temperature features. The weights are dynamically adjusted according to the feature importance (such as geometric feature weight 0.6, spectral feature 0.3, temperature feature 0.1).

[0095] Construct a 3D model for the research and analysis package of tree individual characteristics, including:

[0096] Based on the fused and denoised lidar point cloud data, multispectral images, and thermal infrared image data, use a point cloud-based surface reconstruction algorithm. During the reconstruction process, dynamically adjust the sphere radius parameter according to the local curvature of the tree point cloud. At the same time, map the texture information of the multispectral image to the model surface.

[0097] In specific implementation, point cloud surface reconstruction includes: dynamic sphere radius adjustment, using the moving least squares (MLS) method based on local curvature for surface reconstruction. For areas with high curvature (such as tree branch forks), reduce the sphere radius to 0.1 m to retain details; for flat areas (such as the trunk surface), increase the sphere radius to 0.3 m to improve the reconstruction efficiency. Simplify the reconstructed mesh, use the QuadricEdgeCollapse algorithm to reduce the number of triangles while retaining geometric features, and control the simplification rate at 30% - 40%. In texture mapping, use multispectral texture fusion, that is, map the reflectance data (after radiometric calibration) of the multispectral image to the 3D model surface. For each triangular patch, calculate its average reflectance in the multispectral image to generate a pseudo-color map to highlight the vegetation health status (e.g., red indicates disease, green indicates health).

[0098] Feature parameter extraction for the research and analysis package of tree individual characteristics includes:

[0099] Directly obtain the tree trunk height by measuring the length of the central axis of the tree trunk in the 3D model; calculate the crown shape index based on the 3D shape of the crown; statistically analyze the number of tree branch forks through topological analysis of the branch skeleton structure in the 3D model.

[0100] Tree trunk height measurement is to directly obtain the endpoint coordinates of the central axis of the tree trunk and calculate the Euclidean distance between the two points as the tree trunk height.

[0101] Crown shape index calculation is to calculate the 3D volume and vertical projection area of the crown.

[0102] Tree branch fork number statistics is to perform topological analysis on the skeleton structure, identify nodes with a degree ≥ 3 as fork points, and the fork number is the number of such nodes. Combine the DBSCAN clustering results to distinguish different levels of forks (such as main branch forks, side branch forks).

[0103] Through the above data fusion, 3D modeling, and feature extraction processes, the research and analysis package of tree individual characteristics can accurately analyze the morphology and physiological details of individual trees, providing key technical support for the refined management of forestry resources.

[0104] Fuse two analysis packages and analyze and construct the extracted feature parameters. Specifically, associate the parameters of the two analysis packages through a unified geographic coordinate system, establish an index using a spatial database, configure feature weights according to scenario requirements, construct a macro-micro parameter model, generate a corresponding scenario map, and visualize it through a 3D model, such as Figure 2 as shown

[0105] For the specific implementation of fusing two analysis packages, it includes: unifying the geographic coordinate system, that is, mapping the 3D model (resolution 10m×10m grid) of the macro monitoring analysis package and the single-tree data (accuracy 0.1m level) of the individual research analysis package to the WGS84 coordinate system. For the individual tree point cloud, perform rigid registration of the GPS / IMU positioning data of the lidar point cloud with the ground control points (GCPs) of the macro DSM, and optimize using the iterative closest point (ICP) algorithm to ensure that the single-tree position error < 5cm. Then establish a spatial database index, that is, adopt an R-tree spatial index structure to partition the macro grid and individual trees according to geographical coordinates. For example, take 1km×1km as an index unit, and each unit stores the macro features (such as coverage rate, slope) in the area and the ID and location of individual trees. Establish an association relationship table to record the macro grid ID to which each individual tree belongs and the macro parameters (such as NDVI mean, terrain slope) of this grid

[0106] The dynamic configuration of feature weights is mainly weight allocation driven by scenario requirements. First, define a scenario weight configuration file to support users in dynamically adjusting feature weights according to application goals (such as forest health assessment, resource planning).

[0107] In specific implementation, according to the core goals of forestry surveys, define four typical application scenarios, and allocate weights according to the importance differences between macro features (M) and individual features (I). The feature classification and initial weight configuration are as shown in Table 1 below

[0108] Application scenario Core objective Macroscopic feature (M) Individual feature (I) Basic weight allocation (M:I) Forest health assessment Identify tree diseases and insect pests, growth stress Grid NDVI mean, temperature mean, diurnal temperature difference Single-tree NDVI outliers, crown shape index, thermal infrared temperature variance 0.3:0.7 Carbon storage estimation Regional vegetation biomass and carbon sink capacity assessment Average tree height, vegetation volume, coverage rate Trunk diameter, trunk height, wood density (predicted value) 0.6:0.4 Resource planning (logging / planting) Optimize forest tree spatial distribution and economic value assessment Topographic slope, point cloud density, red edge reflectance DBH (diameter at 1.3 m of the trunk), tree species classification, growth potential index 0.5:0.5 Disaster risk assessment (wind / fire) Identify high-risk areas and tree stress resistance Slope, aspect, vegetation density Crown wind resistance index (based on the number of skeleton branches), thermal infrared hotspot intensity 0.4:0.6

[0109] During implementation, the feature weights for different sub-scenarios can be refined and configured as follows

[0110] In the pest and disease monitoring scenario, the individual feature weight is 0.7 and the macro feature weight is 0.3

[0111] Among them, the weight allocation of the individual feature of 0.7 is as follows: the outlier of the single-tree NDVI (deviation from the mean of healthy trees, weight 0.3), the crown shape index (main canopy volume / vertical projection area, weight 0.25), the variance of the thermal infrared temperature (reflecting the thermal stability of the tree bark, weight 0.15), and the pest and disease mark (based on the classification result of multi-spectral texture features, weight 0.1)

[0112] The weight distribution of the macro features of 0.3 is as follows: the mean of grid NDVI (regional health baseline, weight 0.15), the diurnal temperature difference (environmental stress factor, weight 0.1), and the vegetation density (the number of grid point clouds, weight 0.05).

[0113] If the NDVI outlier of a single tree exceeds 3σ (standard deviation), the weight of the individual feature is automatically increased to 0.8, and the macro feature is reduced to 0.2 to enhance the recognition priority of the abnormal individual.

[0114] In the scenario of carbon storage estimation, the macro feature is 0.6 and the individual feature is 0.4;

[0115] Among them, the weight distribution of the macro feature of 0.6 is as follows: the average tree height (weighted average based on GMM stratification, weight 0.25), the vegetation volume (the volume of tree point clouds in the 3D model, weight 0.2), the coverage rate (the proportion of the volume of tree point clouds in the grid, weight 0.15), and the mean of NDVI (indirect indicator of biomass, weight 0.1);

[0116] The weight distribution of the individual feature (0.4) is as follows: the trunk diameter (diameter at breast height, weight 0.2), the trunk height (the length of the central axis, weight 0.15), and the lidar intensity value (prediction of wood density, weight 0.05).

[0117] For the dense forest area with a canopy density > 0.9, the weight of the macro feature is fine-tuned to 0.65 (increasing the weight of the vegetation volume to 0.25) to reduce the influence of individual feature measurement errors.

[0118] In the resource planning scenario, the macro - individual balanced configuration is (0.5:0.5).

[0119] The weight distribution of the macro feature is as follows: the terrain slope (affecting the logging cost, weight 0.2), the point cloud density (vegetation richness, weight 0.15), and the red-edge reflectance (the growth stage of forest trees, weight 0.15);

[0120] The weight distribution of the individual feature is as follows: the diameter at breast height (core indicator of economic value, weight 0.25), the tree species classification (based on multi-spectral spectral features, weight 0.15), and the growth potential index (combining NDVI and annual increment of tree height, weight 0.1).

[0121] When the user sets "economic value first", the combined weight of "diameter at breast height" and "tree species classification" in the individual feature is increased to 0.4 (original 0.4), and the weight of "red-edge reflectance" in the macro feature is increased to 0.2 (reflecting the maturity of tree species).

[0122] The weight optimization algorithm uses the Analytic Hierarchy Process (AHP) combined with the entropy weight method, combining subjective and objective weighting. The subjective weight is determined by expert scoring, while the objective weight is calculated based on the contribution degree of feature variance. The final weight is the weighted average of the two.

[0123] For the construction of the macro-micro parameter model, first, a fused feature set is extracted, including macro parameters (such as coverage rate, slope, mean NDVI) and individual parameters (such as tree trunk height, crown shape index, pest and disease index). A cross-scale model is constructed using Gradient Boosting Tree (XGBoost) or deep neural networks (such as ResNet), with the fused features as the input and the comprehensive evaluation results (such as forest health level, resource distribution prediction) as the output. During the training process, the model hyperparameters are optimized through cross-validation to ensure the balance between macro trends and individual anomaly recognition. The Apriori algorithm is used to mine the strong association rules between macro and micro features. For example: If the slope of a grid > 30° and the average crown shape index of the trees in this grid < 0.8, then the wind resistance of the trees in this area is low (support > 0.6, confidence > 0.7). If the mean NDVI of a grid < 0.4 and the NDVI of more than 50% of the trees in it < 0.3, then there may be pests and diseases in this area (support > 0.5, confidence > 0.8).

[0124] In the specific implementation, the macro-micro parameter model algorithm uses XGBoost gradient boosting tree, and the parameter configuration is: learning rate 0.1, tree depth 6, subsampling 0.8, column sampling 0.7. The evaluation index is Root Mean Square Error (RMSE) ≤ 0.15. The training data is a fused dataset containing 2000 groups of macro features and 50000 individual features.

[0125] Generate a thematic map of macro-micro fusion, including: a health assessment map, with macro grids as units, superimposing the health indices of individual trees (such as NDVI values, pest and disease marks), and showing the health status distribution with a heat map. A resource planning map, combining macro coverage rate, average tree height and the economic value of individual trees (such as tree species, diameter at breast height), and marking high-value areas. Using the graduated symbol method, key parameters of individual trees are represented by symbols of different sizes and colors on the macro map (for example, the larger the tree trunk diameter, the larger the symbol size). Load the fused model in a 3D GIS platform (such as Cesium, ArcGISPro) to achieve the following functions: multi-scale switching, through zooming operations, smoothly transitioning from the macro forest panorama to the details of individual trees, and real-time displaying the macro and individual parameters at the corresponding positions. Dynamic query, click on any position in the 3D model, and pop up the macro features of this area (such as coverage rate, slope) and the individual features of the nearest 5 trees (such as height, health index). Simulation analysis, based on the fused model, simulate the impact of different management strategies (such as selective logging, pest and disease control) on forest resources, and visualize the result differences.

[0126] The integration of two analysis packages can effectively integrate macro and micro data, provide cross-scale and multi-dimensional decision support for forestry resource management, and achieve the goal of refined management from the global to the local.

[0127] Obviously, the method of the present invention can be implemented by a computer program. The computer program for implementing the method of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, so that when the computer programs are executed by the processor, the functions / operations specified in the flowchart and / or block diagram are implemented. The computer programs can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0128] Therefore, it can be understood that the present invention discloses an electronic device, including:

[0129] At least one processor; and

[0130] A memory communicatively connected to the at least one processor; wherein,

[0131] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned forestry survey planning design and analysis method based on three-dimensional laser modeling.

[0132] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A forestry survey planning and design and analysis method based on three-dimensional laser modeling, characterized in that, It includes the steps of: collecting lidar point cloud data, multispectral images and thermal infrared images of the target area through drones, ground mobile platforms and satellite remote sensing; establishing corresponding analysis packages based on different analysis scenarios, where the analysis packages include an analysis package for the macro monitoring scenario of forest resources and an analysis package for the research scenario of individual tree characteristics. Each analysis package adaptively selects corresponding algorithms for data compression and dimensionality reduction, and then fuses the lidar point cloud data with the multispectral images and thermal infrared images; for each analysis package, constructing a corresponding three-dimensional model using the fused data, and extracting the characteristic parameters required for forestry surveys from the three-dimensional model; fusing the two analysis packages, analyzing and constructing the extracted characteristic parameters, and providing a basis for forestry planning, design and analysis. For the analysis package of the macro monitoring scenario of forest resources, data compression is to perform joint adjustment processing on the lidar point cloud data, multispectral images and thermal infrared images to eliminate redundant data; data dimensionality reduction is to calculate the correlation between the original multi-source data and the macro characteristics of the forest, set the threshold of the correlation coefficient, retain the relevant characteristic data above the threshold, and eliminate the relevant characteristic data below the threshold. Performing joint adjustment processing on the lidar point cloud data, multispectral images and thermal infrared images includes constructing a joint adjustment optimization model. Constructing a joint adjustment optimization model includes: Defining the joint adjustment objective function as: ； Among them, L i , S i , T i are the observed values of the i-th feature points of the lidar point cloud, the multispectral image, and the thermal infrared image, respectively; , , is the optimized theoretical value; α, β, γ are weight allocations. Then, use the Levenberg-Marquardt algorithm for iterative solution to optimize the spatial coordinates of the lidar point cloud, the radiance value of the multispectral image, and the temperature value of the thermal infrared image until the objective function converges. Fusing the two analysis packages, analyzing and constructing the extracted characteristic parameters. Specifically, associate the parameters of the two analysis packages through a unified geographic coordinate system, establish an index using a spatial database, configure the characteristic weights according to the scenario requirements, construct a macro-micro parameter model, generate the corresponding scenario map, and visualize it through a three-dimensional model.

2. The forestry investigation planning design and analysis method based on three-dimensional laser modeling according to claim 1, characterized in that, For the data fusion of the analysis package of the macro monitoring scenario of forest resources, it includes: Denoise the lidar point cloud and extract macro terrain features and vegetation features, convert the pixel DN value of the multispectral image into a radiance value and calculate spectral features, convert the gray value of the thermal infrared image into an actual temperature value and extract energy distribution features; during fusion, first preliminarily align the three types of data according to the geographic coordinates, then perform secondary matching based on feature similarity, adopt weighted fusion, allocate weights according to the importance of the data sources, and fuse the matched feature data according to the allocated weights.

3. A method for forestry survey planning design and analysis based on three-dimensional laser modeling according to claim 2, characterized in that, For constructing a three-dimensional model of the analysis package of the macro monitoring of forest resources, it includes: Use the macro terrain features and vegetation features extracted from the lidar point cloud as the spatial skeleton, endow the model with vegetation details and classification information through the multispectral image, endow the model with energy distribution through the thermal infrared image, construct a three-dimensional model, and render the model surface using the texture and color information of the multispectral image. [[ID= Apply the Sobel operator to the generated digital surface model and calculate the east-west direction gradient G x and the north-south direction gradient G y ; Then calculate the slope, and the slope S has the following calculation formula: ； Aspect calculation is to perform a fast Fourier transform on the generated digital surface model to extract the terrain frequency domain features; the aspect A is calculated by the formula: ; For the high-frequency noise region, recalculate after smoothing with Gaussian filtering.

4. A forestry survey planning, design and analysis method based on three-dimensional laser modeling according to claim 3, characterized in that, ​ The forest coverage rate is obtained by calculating the ratio of the space volume occupied by the tree point cloud in the three-dimensional model to the space volume of the entire target area; the average tree height is calculated by obtaining the height data of the tree point cloud in the three-dimensional model and then using the weighted average method, and the terrain slope is obtained by performing plane fitting analysis on the terrain point cloud in the three-dimensional model.

5. A forestry survey planning, design and analysis method based on three-dimensional laser modeling according to claim 1, characterized in that In the analysis package for the research scenario of tree individual characteristics, data compression is to calculate the covariance matrix of the lidar point cloud data, analyze the variance ratio of each dimension, select the dimension with a large variance ratio as the core segmentation dimension, divide intervals according to the coordinate range of the core segmentation dimension, compress each interval independently, introduce distance weights to divide nodes, cluster the point cloud within each interval based on the weights, identify the core points, retain the core points and their neighboring points, and eliminate low-density noise points; data reduction is to sample and split various features of the lidar point cloud data of a single tree, screen out the key features, set a threshold, and retain the features that exceed the threshold.

6. A forestry survey planning, design and analysis method based on three-dimensional laser modeling according to claim 5, characterized in that, The data fusion of the analysis package for the research scenario of tree individual characteristics includes: Denoise the lidar point cloud data, multi-spectral image, and thermal infrared image of a single tree respectively, then determine the trunk central axis and position by extracting features from the lidar point cloud data of a single tree, then calculate the trunk diameter, distinguish the crown levels, and obtain the branch features; extract texture features from the multi-spectral image of a single tree. When matching, identify the tree contour through the multi-spectral image, then match it with the corresponding lidar point cloud through geometric position consistency, and then associate the thermal infrared and lidar data, and perform weighted fusion of the geometric features of the lidar data, the texture and health index of the multi-spectral, and the temperature features of the thermal infrared.

7. A forestry survey planning, design and analysis method based on three-dimensional laser modeling according to claim 6, characterized in that Construct a three-dimensional model for the analysis package of tree individual characteristics research, including: based on the denoised lidar point cloud data, multi-spectral image, and thermal infrared image data after fusion, use point cloud-based surface reconstruction, dynamically adjust the sphere radius parameter according to the local curvature of the tree point cloud, and at the same time, perform texture mapping on the model surface by combining the texture information of the multi-spectral image.

8. An electronic device, characterized in that, Including: At least one processor; And A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in claim 1.

Images