Calculation method and system for Korean pine leaf area density

Abstract

The invention relates to a Korean pine leaf area density calculation method and system, and belongs to the technical field of laser radar data processing, and the method comprises the steps: carrying out the preprocessing operation of an original point cloud data set of Korean pine to be measured, and obtaining a point cloud data set with the ground surface as a reference; calling a pulse-to-grid algorithm, and converting the point cloud data set taking the earth surface as a reference into a canopy height model for reflecting the spatial height distribution of the Korean pine; performing vertical layering and voxel division based on the canopy height model to obtain a plurality of height layers in the vertical direction and data point number and space density of each layer; according to a plurality of height layers and the data point number and the space density of each layer, the corresponding LAD is calculated layer by layer by using the Beer-Lambert law, and efficient and accurate estimation of the Korean pine leaf area volume density is realized by using laser radar data and combining an advanced data processing algorithm.

Description

Technical Field

[0001] This invention relates to a method and system for calculating the area density of red pine needles, belonging to the field of lidar data processing technology. Background Technology

[0002] As the top-level forest species in Northeast China, Korean pine has a complex canopy structure, with its needles distributed in clusters and exhibiting significant vertical stratification. Accurately estimating its forest floor area (LAD) is of vital scientific and practical value for assessing forest health, monitoring carbon sequestration capacity, and guiding sustainable management.

[0003] Leaf area density (LAD) is the total area of ​​a single leaf surface per unit three-dimensional volume (unit: m³). 2 / m 3 Leaf area index (LAD) is a core physical parameter for quantifying the internal structure and material distribution of vegetation canopy. Compared to the macroscopic leaf area index, LAD can finely characterize the leaf distribution features of the vegetation canopy in the vertical dimension, and is a key input variable for constructing models of forest ecosystem carbon cycling, hydrological processes, biomass estimation, and canopy radiative transfer. Traditional LAD measurement methods mainly rely on destructive field sampling, such as stratified logging or indirect measurements based on optical instruments. These methods are not only time-consuming, labor-intensive, and costly, but also have limited sampling ranges, resulting in inherent defects such as insufficient sample representativeness and difficulty in large-scale, high-time-sensitivity monitoring. With the development of remote sensing technology, LAD inversion using optical imagery (such as multispectral and hyperspectral) and lidar technology has become mainstream. Among them, optical remote sensing indirectly estimates LAD by constructing a statistical relationship between vegetation indices and LAD, but its signals are easily affected by atmospheric conditions, changes in illumination, and soil background reflection. Moreover, it is essentially a two-dimensional observation that cannot penetrate the vegetation canopy, making it difficult to obtain true vertical structure information within the canopy.

[0004] In existing technologies, lidar (LiDAR), as an active remote sensing technology, can acquire high-precision, high-density three-dimensional point cloud data of the ground surface and vegetation canopy by emitting laser pulses and accurately recording the echo signals, providing an unprecedented technical means for the direct and high-precision inversion of LAD. LiDAR technology has strong canopy penetration capabilities, enabling it to directly capture complete vertical structural information from the canopy top to the understory ground cover.

[0005] However, despite its great technological potential, existing LiDAR-based LAD estimation methods still face a series of technical challenges when applied to coniferous forests such as Korean pine, which have complex canopy structures and significant needle aggregation effects. For example, the raw LiDAR point cloud data inevitably contains noise points; the accuracy and robustness of existing algorithms in complex terrain still need to be improved; the physical meaning and adaptability of the model are insufficient; and there are issues with the formation and automation processing systems. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for calculating the area density of red pine needles, in order to solve the problems of poor data processing accuracy, insufficient model adaptability, and inability to automatically obtain LAD. It integrates advanced point cloud preprocessing algorithms with optimized LAD physical inversion models, aiming to promote LAD estimation technology from theoretical research to standardization and business application.

[0007] To achieve the above objectives, this invention proposes a method for calculating the area density of red pine needles, comprising the following steps:

[0008] 1) Perform preprocessing operations on the original point cloud dataset of the red pine to be measured to obtain a point cloud dataset based on the ground surface;

[0009] 2) Use the pulse-to-raster algorithm to convert the point cloud dataset based on the ground surface into a canopy height model that reflects the spatial height distribution of the Korean pine.

[0010] 3) Based on the canopy height model, vertical stratification and voxel division are performed to obtain several height layers in the vertical direction and the number of data points and spatial density of each layer; according to the several height layers and the number of data points and spatial density of each layer, the corresponding LAD is calculated layer by layer using the Beer-Lamber law.

[0011] Furthermore, based on several height layers and the number of data points and spatial density in each layer, the Beer-Lamber law is used to calculate the LAD of each layer using the following formula:

[0012] ;

[0013] in, Let LAD be the value of the i-th layer; This represents the point cloud density at the top of the canopy (unobstructed area); Let be the spatial density of the i-th layer; k is an empirical coefficient.

[0014] Furthermore, based on the canopy height model, vertical stratification and voxel division are performed, and the following methods are used to obtain several height layers in the vertical direction, as well as the number of data points and spatial density of each layer:

[0015] In three-dimensional space, the canopy height model is divided into several height layers according to a preset unit cube grid, and the height range of each height layer is calculated.

[0016] Based on the height range of each height layer, the number of data points and spatial density of each height layer are calculated.

[0017] Furthermore, the height range of each height layer is expressed by the following formula:

[0018]

[0019] in, dz represents the height range of the i-th layer; dz represents the vertical resolution. This represents the minimum height range.

[0020] Furthermore, when calling the pulse-to-raster algorithm, each pulse in the point cloud dataset based on the ground surface is converted into a raster cell, and the height value of each raster cell is the highest elevation value of all point cloud data in that cell, thereby generating the canopy height model of the red pine.

[0021] Furthermore, the preprocessing operation includes noise removal. During noise removal, a statistical discrete cluster removal algorithm and an isolated point voxel filtering algorithm are invoked to eliminate noise points in the original point cloud dataset; wherein,

[0022] The statistical discrete cluster removal algorithm is used to remove outliers from all points in the original point cloud dataset.

[0023] The isolated point voxel filtering algorithm is used to divide the point cloud space corresponding to the original point cloud dataset into a three-dimensional voxel grid of a preset fixed size; by counting the number of points in each voxel grid, all points in voxel grids with a number of points lower than a preset point number threshold are marked as noise points and removed.

[0024] Furthermore, the preprocessing operation also includes ground point classification. During ground point classification, a progressive morphological filtering algorithm and a cloth simulation filtering algorithm are used to separate the denoised original point cloud dataset into ground points and non-ground points; wherein,

[0025] The progressive morphological filtering algorithm performs morphological opening operations on the denoised original point cloud dataset through a dynamically increasing window, thereby separating non-ground points from the denoised original point cloud dataset.

[0026] The fabric simulation filtering algorithm is used to identify ground points in the denoised original point cloud dataset.

[0027] Furthermore, the preprocessing operation also includes point cloud elevation value normalization. During this normalization process, the relative ground height of vegetation is calculated using an irregular triangular mesh interpolation algorithm and a height normalization algorithm to obtain a point cloud dataset based on the land surface.

[0028] Using the divided ground point dataset, a raster digital elevation model is generated through an irregular triangular mesh interpolation algorithm;

[0029] The height normalization algorithm is used to subtract the elevation value in the digital elevation model of each point in the divided ground point dataset from the original elevation value of that point to obtain the relative ground height of that point; based on the relative ground heights of all points in the divided ground point dataset, the point cloud dataset with the ground surface as the reference is obtained.

[0030] Furthermore, it also includes: calculating the leaf area index using the following formula:

[0031]

[0032] LAI stands for Leaf Area Index. Let LAD be the layer i. is the thickness of the i-th layer; n is the total number of layers.

[0033] On the other hand, the present invention also proposes a system for calculating the area density of red pine needles, including a processor, the processor being used to perform the following method steps:

[0034] 1) Perform preprocessing operations on the original point cloud dataset of the red pine to be measured to obtain a point cloud dataset based on the ground surface;

[0035] 2) Use the pulse-to-raster algorithm to convert the point cloud dataset based on the ground surface into a canopy height model that reflects the spatial height distribution of the Korean pine.

[0036] 3) Based on the canopy height model, vertical stratification and voxel division are performed to obtain several height layers in the vertical direction and the number of data points and spatial density of each layer; according to the several height layers and the number of data points and spatial density of each layer, the corresponding LAD is calculated layer by layer using the Beer-Lamber law.

[0037] Furthermore, based on several height layers and the number of data points and spatial density in each layer, the Beer-Lamber law is used to calculate the LAD of each layer using the following formula:

[0038] ;

[0039] in, Let LAD be the value of the i-th layer; This represents the point cloud density at the top of the canopy (unobstructed area); Let be the spatial density of the i-th layer; k is an empirical coefficient.

[0040] Furthermore, based on the canopy height model, vertical stratification and voxel division are performed, and the following methods are used to obtain several height layers in the vertical direction, as well as the number of data points and spatial density of each layer:

[0041] In three-dimensional space, the canopy height model is divided into several height layers according to a preset unit cube grid, and the height range of each height layer is calculated.

[0042] Based on the height range of each height layer, the number of data points and spatial density of each height layer are calculated.

[0043] Furthermore, the height range of each height layer is expressed by the following formula:

[0044]

[0045] in, dz represents the height range of the i-th layer; dz represents the vertical resolution. This represents the minimum height range.

[0046] Furthermore, when calling the pulse-to-raster algorithm, each pulse in the point cloud dataset based on the ground surface is converted into a raster cell, and the height value of each raster cell is the highest elevation value of all point cloud data in that cell, thereby generating the canopy height model of the red pine.

[0047] Furthermore, the preprocessing operation includes noise removal. During noise removal, a statistical discrete cluster removal algorithm and an isolated point voxel filtering algorithm are invoked to eliminate noise points in the original point cloud dataset; wherein,

[0048] The statistical discrete cluster removal algorithm is used to remove outliers from all points in the original point cloud dataset.

[0049] The isolated point voxel filtering algorithm is used to divide the point cloud space corresponding to the original point cloud dataset into a three-dimensional voxel grid of a preset fixed size; by counting the number of points in each voxel grid, all points in voxel grids with a number of points lower than a preset point number threshold are marked as noise points and removed.

[0050] Furthermore, the preprocessing operation also includes ground point classification. During ground point classification, a progressive morphological filtering algorithm and a cloth simulation filtering algorithm are used to separate the denoised original point cloud dataset into ground points and non-ground points; wherein,

[0051] The progressive morphological filtering algorithm performs morphological opening operations on the denoised original point cloud dataset through a dynamically increasing window, thereby separating non-ground points from the denoised original point cloud dataset.

[0052] The fabric simulation filtering algorithm is used to identify ground points in the denoised original point cloud dataset.

[0053] Furthermore, the preprocessing operation also includes point cloud elevation value normalization. During this normalization process, the relative ground height of vegetation is calculated using an irregular triangular mesh interpolation algorithm and a height normalization algorithm to obtain a point cloud dataset based on the land surface.

[0054] Using the divided ground point dataset, a raster digital elevation model is generated through an irregular triangular mesh interpolation algorithm;

[0055] The height normalization algorithm is used to subtract the elevation value in the digital elevation model of each point in the divided ground point dataset from the original elevation value of that point to obtain the relative ground height of that point; based on the relative ground heights of all points in the divided ground point dataset, the point cloud dataset with the ground surface as the reference is obtained.

[0056] Furthermore, it also includes: calculating the leaf area index using the following formula:

[0057]

[0058] LAI stands for Leaf Area Index. Let LAD be the layer i. is the thickness of the i-th layer; n is the total number of layers.

[0059] The beneficial effects of this invention are as follows: Preprocessing is performed on the original point cloud dataset of the Korean pine to be measured to obtain a point cloud dataset based on the ground surface; the pulse-to-raster algorithm is called to convert the point cloud dataset based on the ground surface into a canopy height model that reflects the spatial height distribution of the Korean pine; vertical stratification and voxel division are performed based on the canopy height model to obtain several height layers in the vertical direction, as well as the number of data points and spatial density of each layer; according to the several height layers and the number of data points and spatial density of each layer, the Beer-Lambert law is used to calculate the corresponding LAD layer by layer, thereby removing noise from the original point cloud dataset, accurately separating ground points from non-ground points, and eliminating the influence of terrain undulations on the absolute height of vegetation, resulting in a pure vegetation point cloud based on the ground surface. Attached Figure Description

[0060] Figure 1 This is a flowchart of a method for calculating the area density of red pine needles proposed in this invention;

[0061] Figure 2 This is a schematic diagram of the spatial distribution of a digital terrain model (DTM) generated in the Wuying Nature Reserve in Yichun, Heilongjiang Province, using the method for calculating the area density of red pine needles proposed in this invention.

[0062] Figure 3This is a schematic diagram of the spatial distribution of the canopy height model (CHM) generated based on lidar data in the Wuying Nature Reserve in Yichun, Heilongjiang Province, using a method for calculating the area density of red pine needles proposed in this invention.

[0063] Figure 4 This is a schematic diagram of the spatial distribution of leaf area index (LAI) generated based on lidar data in the Wuying Nature Reserve in Yichun, Heilongjiang Province, using a method for calculating the leaf area density of red pine proposed in this invention.

[0064] Figure 5 This is a schematic diagram of the spatial distribution of leaf area density (LAD) generated based on lidar data in the Wuying Nature Reserve in Yichun, Heilongjiang Province, using a method for calculating the leaf area density of red pine proposed in this invention.

[0065] Figure 6 This is a schematic diagram of the LAI (Lower Area Area) contour calculated in practical application of the method for calculating the area density of red pine needles proposed in this invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0067] The inventive concept of this invention is to systematically address the shortcomings of existing technologies in accurately and efficiently estimating the area density (LAD) of complex coniferous forests. This invention provides an integrated, high-precision method and system for calculating the area density of red pine needles, integrating point cloud preprocessing algorithms with an optimized LAD physical inversion model, thereby promoting the standardization and operationalization of LAD estimation technology from theoretical research.

[0068] Detailed implementation method 1:

[0069] like Figure 1 The diagram shows a flowchart of a method for calculating the area density of red pine needles proposed in this invention, which includes steps S11, S12, and S13, specifically:

[0070] Step S11 involves preprocessing the original point cloud dataset of the Korean pine to be measured to obtain a point cloud dataset based on the land surface. Here, the original point cloud dataset of the Korean pine to be measured is a high-precision, high-density three-dimensional point cloud dataset obtained through lidar technology to represent the land surface and vegetation canopy. During the preprocessing operation, noise removal, precise separation of ground points and non-ground points, and elimination of the influence of terrain undulations on the absolute height of vegetation are performed to obtain a clean vegetation point cloud based on the land surface.

[0071] Step S12: Invoke the pulse-to-raster algorithm to convert the point cloud dataset based on the ground surface into a canopy height model that reflects the spatial height distribution of the Korean pine. The pulse-to-raster algorithm refers to p2r (Pulse to Raster). When the pulse-to-raster algorithm is invoked, each pulse in the point cloud dataset based on the ground surface is converted into a raster cell, and the height value of each raster cell is the highest elevation value of all point cloud data in that cell, thereby generating the canopy height model of the Korean pine.

[0072] Step S13: Perform vertical stratification and voxel division based on the canopy height model to obtain several height layers in the vertical direction and the number of data points and spatial density of each layer; according to the several height layers and the number of data points and spatial density of each layer, use the Beer-Lamber law to calculate the corresponding LAD layer by layer.

[0073] Through steps S11-S13, efficient, accurate, and reproducible estimation of the three-dimensional structural parameters of coniferous vegetation such as Korean pine is achieved, providing key technical support and data foundation for refined management of forest resources, ecosystem function assessment, and regional carbon sink research.

[0074] Method Detailed Implementation 2:

[0075] The following section explains the method for calculating the area density of red pine needles proposed in this invention, using practical application scenarios as examples.

[0076] The data preprocessing step is fundamental to ensuring the accuracy of LAD calculations. This step aims to remove noise from the raw point cloud, accurately separate ground points from non-ground points, and eliminate the influence of terrain undulations on the absolute height of vegetation, resulting in a clean, surface-based vegetation point cloud. This step includes noise removal, ground point classification, and point cloud height normalization, specifically:

[0077] 1. Noise Removal: To eliminate noise points (such as floating points in the air) introduced by sensor measurement errors, atmospheric particulate matter reflection, or multipath effects, a composite strategy combining Statistical Outlier Removal (SOR) and Isolated Voxel Filter (IVF) is adopted.

[0078] Statistical Outlier Removal (SOR) is based on the statistical properties of local neighborhoods in a point cloud. Its core assumption is that the neighborhood density of real-world points is relatively uniform, while noise points deviate significantly from this distribution. The algorithm calculates the average distance d(p) from each point p to its k nearest neighbors. Then, it calculates the global mean μ and standard deviation σ of the average distances to all points in the entire point cloud. A point p is identified as an outlier and removed if its d(p) satisfies the following condition:

[0079] ;

[0080] Where k (number of neighborhood points) and (Standard deviation factor) is an adjustable parameter used to control the stringency of the filter. This method can effectively remove sparse, statistically anomalous noise points.

[0081] Isolated point voxel filtering (IVF) was introduced to address the shortcomings of the SOR algorithm in removing small-scale noise clumps. This algorithm divides the point cloud space into a fixed-size 3D voxel grid. By counting the number of points within each voxel, all points within voxels with a count below a preset threshold are marked as noise points and removed. This method can efficiently remove noise clumps that, although their local density is not zero, are macroscopically isolated.

[0082] 2. Ground Point Classification: Accurate separation of ground points is crucial for subsequent generation of the Digital Terrain Model (DTM) and high-level normalization. A progressive morphological filter (PMF) and a cloth simulation filter (CSF) are used in synergistic processing to address the complex and varied terrain under the Korean pine forest.

[0083] Progressive Morphological Filtering (PMF) takes advantage of the relatively gentle elevation changes of ground points. It performs morphological opening operations (erosion followed by dilation) on the point cloud through a series of windows of progressively increasing size, gradually separating non-ground points (such as vegetation and buildings) from the point cloud. This method performs well in areas with relatively flat terrain.

[0084] To overcome the shortcomings of PMF (Profile Filtering) in complex terrains such as steep slopes and ditches, CSF (Cloth Simulation Filtering) is introduced as a supplement and optimization. CSF identifies ground points by simulating the physical process of a rigid virtual "cloth" falling under gravity and conforming to the terrain surface above an inverted point cloud. Points that finally reach equilibrium with the cloth, in contact with or close to the cloth, are classified as ground points. CSF exhibits excellent adaptability to complex and steep terrains, and studies show that its overall performance is generally superior to PMF. The combination of the two enables robust classification of various terrain types.

[0085] 3. Point Cloud Height Normalization: To eliminate the influence of terrain undulations on the absolute height of vegetation, height normalization is required to calculate the relative ground height of vegetation. Digital Terrain Model (DTM) Generation: Using the ground points classified in the previous step, a high-resolution raster DTM is generated through the Triangulated Irregular Network (TIN) interpolation method. TIN can preserve the detailed features of the terrain to the greatest extent by constructing a network of triangular facets connecting all ground points, and is especially suitable for areas with complex terrain. Height Normalization: The original elevation value (Z_raw) of each point in the point cloud is subtracted from the DTM elevation value (Z_DTM) corresponding to its plane position (X, Y) to obtain the normalized height (Z_normalized = Z_raw - Z_DTM). The normalized point cloud accurately reflects the relative ground height of vegetation and ground features, and is the basis for all subsequent calculations of canopy structure parameters.

[0086] The process involves model building and LAD calculation, which includes generating the Canopy Height Model (CHM), calculating LAD, and calculating LAI (Leaf Area Index). Specifically:

[0087] 1. Canopy Height Model (CHM) Generation: The CHM visually displays the spatial height distribution of the vegetation canopy. It is generated using normalized non-ground point cloud data and a pulse-to-raster (p2r) algorithm. This algorithm assigns the value of each raster cell to the highest elevation value of all point clouds within that cell, accurately preserving detailed canopy information such as the highest point of the canopy and the location of forest gaps.

[0088] 2. LAD Calculation: LAD calculation is the core process, based on vertical hierarchical statistics of point clouds and inversion using the Beer-Lambert physical model. 3D Voxelization: Normalized point cloud data is divided into regular cubic meshes, or voxels, in 3D space. Voxelization is the process of discretizing a continuous point cloud space into standard units, facilitating spatial statistics and analysis. The size of the voxels (e.g., 0.5m × 0.5m × 0.5m) is a key parameter, requiring a trade-off between point cloud density and the research scale. Too large a size results in loss of detail, while too small a size leads to enormous computational costs and susceptibility to noise.

[0089] LAD estimation model (based on Beer-Lambert's law): The core theoretical basis is Beer-Lambert's law, which was originally used to describe the intensity attenuation of light as it propagates through an absorbing medium. In LiDAR applications, this law has been creatively extended, analogizing the penetration process of a laser pulse through a vegetation canopy to the attenuation process of light in a medium. For the i-th height layer in the vertical direction (consisting of a single voxel), its LAD value (LAD...) is... i It can be estimated using the following formula:

[0090] ;

[0091] in, denoted by , k represents the point cloud density at the top of the canopy (unobstructed area); k is the extinction coefficient, a dimensionless parameter describing the extinction capacity of a unit leaf area for laser light. It is related to factors such as the leaf tilt angle distribution, aggregation effect, and the scanning angle of the LiDAR. The k value usually needs to be calibrated through field measurements or references based on the vegetation type and LiDAR system parameters of the study area. For coniferous forests, the k value is usually around 0.5. This formula calculates the laser penetration rate within each canopy layer, thus reversing the leaf area density of that layer, and has a clear physical meaning.

[0092] 3. LAI Calculation: Leaf Area Index (LAI) is defined as the total leaf area per unit area of ​​the land surface, and can be obtained by integrating the leaf area index (LAD) in the vertical direction. The formula for calculating LAI is as follows:

[0093] ;

[0094] After discretization, it is represented as:

[0095] ;

[0096] in, Let be the thickness of the i-th layer.

[0097] To eliminate interference from low understory vegetation or ground echo noise, the integration is typically calculated starting from a preset lower limit of the canopy height and continuing up to the top of the canopy. By performing independent vertical LAD integration on each horizontal grid cell, a spatially continuous LAI distribution map can be generated.

[0098] Method Detailed Implementation 3:

[0099] The following section explains the method for calculating the area density of Korean pine needles proposed in this invention, using a practical application scenario. It should be noted that the research area of ​​this invention must ensure that Korean pine is widely distributed and forms a typical coniferous forest ecosystem. Therefore, in this embodiment, the preferred research area is the Wuying Nature Reserve in Yichun, Heilongjiang Province.

[0100] First, point cloud data of the study area was acquired using an airborne LiDAR system in 2024. The LiDAR system emitted a high-frequency laser beam and precisely recorded the round-trip time between the laser and the target object, obtaining detailed three-dimensional point cloud data of the target area.

[0101] Next, data preprocessing operations are performed, including noise removal, ground point classification, and point cloud normalization. Specifically:

[0102] (1) Noise removal includes the Statistical Outlier Removal (SOR) algorithm and the Instantaneous Filtering (IVF) algorithm:

[0103] Point cloud data is read using the professional point cloud processing tool CloudCompare. This software provides visualized point clouds for a clear understanding of data distribution and supports custom point sizes and background colors. First, the point cloud is checked for coordinate system information; if missing, a warning is issued prompting the user to add a coordinate reference system (such as WGS 84 / UTM). Second, the point cloud undergoes preliminary filtering to remove obvious noise points. Finally, it is crucial to ensure the integrity of the point cloud data to avoid subsequent processing failures due to file corruption.

[0104] SOR (Statistical Outlier Removal) is based on statistical analysis of the local neighborhood of point cloud. It assumes that the elevation distribution of normal points follows a Gaussian distribution, while the distribution of outliers (noise) deviates from the main data [4]. The algorithm calculates the average distance of each point to its k nearest neighbors and removes points whose distance exceeds the average distance of its k nearest neighbors. ( The mean; Standard deviation; Points that are multiples of the threshold.

[0105] Formula parameter: k is the number of neighborhood points (default k=10), which determines the scope of local statistics. Multiples of standard deviation (default) =3), controls the threshold for outlier detection. Function: The larger the value of k, the more global the statistical analysis; The smaller the value, the more stringent the noise reduction requirements.

[0106] Isolated Voxel Filter (IVF) is a voxel-based point cloud denoising method. It divides the point cloud into a fixed-size voxel grid and identifies noise points based on the number of points within each voxel. The specific steps are as follows: 1. Voxel grid division: Divide the point cloud data into a cubic voxel grid with a variable length of 1 meter. 2. Voxel point count: Count the number of points within each voxel. 3. Noise point labeling: Label points with fewer than 5 points within a voxel as noise points and set their classification value to 18. 4. Noise point removal: Remove the labeled noise points from the point cloud.

[0107] (2) Ground point classification: Progressive Morphological Filtering (PMF) and Cloth Simulation Filtering (CSF):

[0108] PMF (Progressive Morphological Filter) progressively separates ground points through a multi-scale sliding window, making it suitable for scenarios with significant terrain undulations. Its core idea is that ground points exhibit gradual elevation changes, while non-ground points (such as vegetation) show abrupt elevation changes. The algorithm steps are as follows: 1. Window initialization: Start with a small window (e.g., 1 m) and gradually increase it to a preset value (e.g., 3 m). 2. Elevation difference calculation: Calculate the elevation difference of the point cloud within the window. 3. Threshold determination: If If the value is less than the threshold (th = 0.5 m), it is marked as a ground point; otherwise, it is discarded.

[0109] Formula parameters: ws: window size (default 3 m), determines the sensitivity to terrain undulation. th: elevation difference threshold (default 0.5 m), controls the strictness of ground point classification.

[0110] CSF (Cloth Simulation Filter) is a LiDAR point cloud ground filtering and segmentation method based on cloth simulation, designed to solve the problem of traditional methods struggling to accurately distinguish between ground and non-ground points in complex terrain. CSF overlays a virtual cloth onto an inverted point cloud, simulating the cloth's descent to identify ground points. This method adapts well to various terrain features, including steep slopes, cliffs, and complex urban environments. The steps are as follows: 1. Point Cloud Inversion: Invert the original point cloud so that ground points are on top. 2. Cloth Initialization: Generate a virtual cloth on top of the inverted point cloud. 3. Cloth Descent Simulation: Simulate the cloth's descent under gravity, guiding it to contact the point cloud and reach a stable state. 4. Ground Point Identification: Points in contact with or very close to the cloth are identified as ground points; other points are identified as non-ground points.

[0111] (3) Point cloud normalization: TIN interpolation to generate DEM:

[0112] TIN (Triangulated Irregular Network) fits the terrain surface using an irregular triangular mesh, where each triangle vertex represents a ground point. Its core idea is to connect the ground point cloud into triangular patches, with each triangle vertex representing the elevation value of a ground point. DEM (Digital Elevation Model, representing a bare ground surface after removing all natural and artificial features) generates raster elevation data through TIN interpolation. Point cloud normalization involves subtracting the DEM value from the original elevation to obtain the relative vegetation height (Z). normalized = Z raw -Z DEM Parameters and advantages: res: raster resolution (10 m), affecting DEM detail and computational efficiency. Advantages: TIN preserves terrain detail and is suitable for complex terrain areas.

[0113] Continuing, let's discuss Digital Terrain Model (DTM) generation and height normalization. A DTM is a terrain surface model generated by interpolating point cloud data. It describes the elevation information of the ground. The point cloud data refers to denoised and classified point cloud data—that is, point cloud data after noise removal and ground point classification. Height normalization normalizes the height values ​​of the point cloud data onto the DTM to eliminate the influence of terrain undulations on the point cloud data. It's important to note that DTMs typically enhance DEMs by incorporating natural terrain vector features (such as rivers and ridges). A DTM can interpolate to generate a DEM, but the reverse is not true.

[0114] TIN (Triangulated Irregular Network) fits the terrain surface using an irregular triangular network, where each triangle vertex represents a ground point. Its core idea is to connect the ground point cloud into triangular patches, with each triangle vertex representing the elevation value of a ground point. The steps are as follows: 1. Read point cloud data: Use `lidR::readLAS` to read the classified point cloud data. 2. Interpolate to generate a DTM: Generate a DTM using the `lidR::rasterize_terrain` and `lidR::tin` functions, setting the resolution `res` (10 meters). 3. Save the DTM: Save the generated DTM as a GeoTIFF file. 4. Height normalization: Use the `lidR::normalize_height` function to normalize the height values ​​of the point cloud data onto the DTM, and save the normalized point cloud data.

[0115] Based on the generated DTM, a canopy height model (CHM) is produced: The CHM (Canopy Height Model) is a canopy height model generated by rasterizing normalized point cloud data, used to describe the height information of the vegetation canopy. CHM has significant application value in forestry surveys, urban greening management, and other fields.

[0116] The p2r (Pulse to Raster) method converts each pulse in the point cloud data into a raster cell, with the height value of each raster cell determined by the highest point within that cell. This method can better preserve the detailed information of the canopy and is suitable for high-precision canopy height extraction. The steps are as follows: 1. Read point cloud data: Use `lidR::readLAS` to read the normalized point cloud data. 2. Filter negative Z-value points: Filter out points with negative Z-values ​​to ensure data validity. 3. Rasterize and generate CHM: Generate CHM using the `lidR::rasterize_canopy` and `lidR::p2r` functions, setting the resolution (10 meters) and radius (10 meters). 4. Save CHM: Save the generated CHM as a GeoTIFF file. 5. Smoothing (optional): If smoothing is required, use the `terra::focal` function to filter the CHM, setting the kernel size (default is 3) and filter type (median). 6. Visualization: Use the lidR::plot function to plot the CHM and save it as an image file.

[0117] Finally, the calculation process and principle of LAD: LAD (Leaf Area Density) is an important parameter describing the structure of vegetation canopy, representing the leaf area per unit volume. LAD calculation is based on the hierarchical statistics of point cloud data. By analyzing the point density at each height layer, it reflects the leaf area distribution of the canopy at different heights.

[0118] The formula for calculating LAD is as follows: Where LAD(z) is the leaf area density at height z; A(z) is the leaf area (m²) at height z. 2 V(z) is the volume (m³) at height z. 3 In lidar data, leaf area A(z) can be estimated from the height distribution and density of the point cloud data, while volume V(z) is calculated from the volume of each height layer. In actual calculations, LAD is estimated using the vertical distribution information of the point cloud. Assuming that vegetation leaves follow a specific distribution pattern in the vertical direction, the point cloud is divided into a three-dimensional mesh based on a voxelization method, and the leaf area within each voxel is calculated. Specifically:

[0119] Vertical layering and voxel division: Point clouds are layered by height, with a layer thickness of [missing information]. Each layer is considered a horizontal sheet. Assuming the vertical resolution is dz, the height range of the i-th layer is:

[0120] ;

[0121] Count the number of points N in each layer i And calculate its spatial density. : .

[0122] LAD estimation model: Based on Beer-Lambert's law and point cloud density, calculate the LAD value:

[0123] ;

[0124] in, denoted as point cloud density at the top of the canopy (unobstructed area); k: an empirical coefficient related to vegetation type and lidar parameters. This formula infers leaf area density from laser penetration rate, and the value of k is determined through literature review.

[0125] The calculation steps can be summarized as follows: Layered statistics: Divide the point cloud according to the vertical resolution dz, and count the number of points N in each layer. i Density calculation: Calculate the density of each layer. and total density LAD Inversion: Calculate the LAD value layer by layer using the above formula; Output: Save as a CSV file, containing the height z and the corresponding LAD value.

[0126] Based on the LAD calculation results, LAI can be further calculated. The specific process and principle are as follows:

[0127] LAI (Leaf Area Index) is the total leaf area per unit land surface area and is an important indicator for evaluating forest ecosystems. LAI retrieval is based on high-precision three-dimensional information from lidar data, and is calculated through steps such as voxelization, canopy height correction, and LAI estimation.

[0128] The formula for calculating LAI is as follows:

[0129] ;

[0130] After discretization, it is represented as:

[0131] ;

[0132] in, Let be the thickness of the i-th layer.

[0133] The above describes the vertical integration of LAD: LAI is achieved through the vertical integration of LAD, with the discretization formula as shown above. Canopy height definition: Only canopy height is counted. For LAD values ​​above 0.5, ignore low vegetation or ground noise. Set a height threshold. =5 meters, then: .

[0134] Based on the LAI calculation results, spatial rasterization can be performed: to generate a spatially distributed LAI map, the point cloud needs to be rasterized according to the horizontal resolution ( The area is divided into raster cells. The LAI of each raster is calculated independently: Voxel partitioning: The horizontal plane is divided into... The raster; vertical layering: layering the point cloud within each raster by dz; grid-by-grid integration: calculating the LAI value of each raster.

[0135] Method Detailed Implementation 4:

[0136] Following the specific embodiments described above, and referring to the accompanying drawings, the present invention will be further described. Figure 2-6 The method for calculating the area density of red pine needles proposed in this invention will be further explained.

[0137] like Figure 2 The image shows a spatial distribution diagram of a digital terrain model (DTM) generated in the Wuying Nature Reserve in Yichun, Heilongjiang Province, using the method for calculating the area density of red pine needles proposed in this invention. The resolution is 10 meters. This model reflects the elevation characteristics of the surface topography in the study area, eliminating the influence of non-topographic elements such as vegetation and buildings, and providing basic data support for topographic analysis and ecological research.

[0138] Topographical and Elevation Distribution Characteristics: The DTM shows significant elevation differences within the region, exhibiting an overall mountainous and hilly terrain. The highest point is located in the northeast, at an elevation of 505.1 meters, indicating that this area may consist of steep ridges or peaks with complex terrain and large slopes. Lower elevation areas are concentrated in the southwest, with the lowest elevation at 348.0 meters, presumably formed by valleys or river erosion, with relatively gentle terrain and possibly streams or wetlands. The central region has elevations between 400 and 450 meters, with terrain mainly consisting of gentle slopes and hilly transition zones, suitable for vegetation growth and soil development.

[0139] Spatial heterogeneity analysis: Over a horizontal range of approximately 0.36 miles (about 579 meters) covering a longitude range (129° 1' 30° E to 129° 1' 0° E), the topographic elevation varies significantly over short distances, with local elevation differences exceeding 150 meters, reflecting the typical low-to-medium mountain topography of the protected area. High- and low-value areas are interspersed, forming distinct micro-topographic units that may significantly influence local hydrological processes (such as runoff collection) and vegetation differentiation (such as shady and sunny slopes).

[0140] Ecological significance: DTM elevation data provides a basis for analyzing the relationship between topography and vegetation distribution. High-altitude areas (>450 meters) may develop cold-resistant shrubs or coniferous forests due to lower temperatures and stronger winds; low-altitude areas (<380 meters) may have broad-leaved forests or wetland vegetation due to favorable water conditions. In addition, steep slope areas (with large elevation changes) may face a higher risk of soil erosion and require special attention in ecological protection.

[0141] like Figure 3 The image shows a spatial distribution diagram of the canopy height model (CHM) generated based on lidar data in the Wuying Nature Reserve, Yichun City, Heilongjiang Province, using the method for calculating the area density of red pine needles proposed in this invention. The resolution is 0.5 meters. This model reflects detailed information on the vegetation canopy height within the study area, providing important data support for vegetation structure analysis and ecosystem research.

[0142] Canopy height distribution characteristics: CHM shows significant differences in vegetation height within the region, exhibiting an overall vertical structure characteristic of forest canopies. The highest canopy heights are found in the central and southeastern parts, with some areas reaching 83.0 meters, indicating that these areas are mainly composed of tall trees, possibly mature coniferous or broad-leaved forests. Low-value areas are mainly distributed in the edge zones and some open areas, with canopy heights mostly below 10 meters, and the lowest values ​​approaching 0 meters, presumably forest clearings, shrublands, or areas covered by low vegetation.

[0143] Spatial heterogeneity analysis: Over a horizontal range of approximately 0.36 miles (about 579 meters) across the longitude range (129° 10' 30" E to 129° 11' 30" E), canopy height varies significantly over short distances, with local elevation differences exceeding 40 meters, reflecting the complexity and diversity of the forest canopy within the reserve. High- and low-value areas are interspersed, forming distinct vegetation patches that may have a significant impact on local microclimates (such as temperature and humidity) and biodiversity (such as bird habitats).

[0144] Ecological significance: Canopy height data from the CHM provide a basis for analyzing the relationship between vegetation structure and ecological function. High canopy areas (>40 meters) may develop tall trees due to ample sunlight and fertile soil, forming complex vertical structures and providing habitats for various wildlife. Low canopy areas (<10 meters) may be characterized by insufficient sunlight or poor soil, with low shrubs or herbaceous plants. Furthermore, the vertical stratification of canopy height may affect forest productivity and carbon storage capacity.

[0145] Technical application value: The 0.5-meter high-resolution CHM can accurately depict the three-dimensional structure of vegetation canopy, supporting forest biomass estimation, canopy gap analysis and habitat suitability assessment, and providing key data support for forest resource management, biodiversity conservation and ecological restoration planning in protected areas.

[0146] like Figure 4 The image shows a spatial distribution diagram of leaf area index (LAI) generated based on lidar data in the Wuying Nature Reserve, Yichun City, Heilongjiang Province, using a method for calculating the leaf area density of Korean pine proposed in this invention. The resolution is 0.5 meters. This model reflects the total leaf area per unit surface area in the study area, providing important data support for assessing the photosynthetic capacity and ecological function of vegetation.

[0147] Leaf Area Index (LAI) Distribution Characteristics: The LAI shows significant differences in leaf area index within the region, generally exhibiting the leaf cover characteristics of a forest canopy. High-value areas are mainly concentrated in the central and southeastern parts, with LAI values ​​reaching 12.1, indicating dense leaf cover in these areas, possibly concentrated distribution areas of tall trees with strong photosynthetic capacity. Low-value areas are mainly distributed in the marginal zones and some open areas, with LAI values ​​mostly below 2, presumably areas of low shrubs or sparse vegetation cover.

[0148] Spatial heterogeneity analysis: Over a horizontal range of approximately 0.36 miles (about 579 meters) across the longitude range (129° 10' 30" E to 129° 11' 30" E), leaf area index (LAI) showed significant variations over short distances, with local differences exceeding 10, reflecting the complexity and diversity of vegetation cover within the protected area. High and low LAI zones were interspersed, forming a distinct cover gradient, which may have a significant impact on local microclimates (such as temperature and humidity) and soil moisture (such as interception and transpiration).

[0149] Ecological significance: LAI index data provides a basis for analyzing the relationship between vegetation cover and ecological function. High index areas (>8) may become core areas of forest productivity due to dense leaf cover and high photosynthetic efficiency, and are also major contributors to carbon sequestration. Low index areas (<2) may have sparse leaf cover and better light conditions, with low shrubs or herbaceous plants distributed there. In addition, the spatial distribution of leaf area index may affect the forest's water and nutrient cycles.

[0150] like Figure 5 The image shows a schematic diagram of the spatial distribution of leaf area density (LAD) generated based on lidar data in the Wuying Nature Reserve, Yichun City, Heilongjiang Province, using a method for calculating the leaf area density of Korean pine proposed in this invention. The resolution is 0.5 meters. This model reflects the vertical density distribution of vegetation leaf area within the study area, providing important data support for understanding the photosynthetic and energy exchange processes of vegetation.

[0151] Leaf area density distribution characteristics: The LAD (Leaf Area Density Analysis) showed significant differences in leaf area density within the region, generally exhibiting the density variation characteristics of a forest canopy. High-value areas were mainly concentrated in the central and southeastern parts, with leaf area densities reaching 48 / m², indicating that these areas had dense vegetation and were likely the top canopies of tall trees with strong photosynthetic capacity. Low-value areas were mainly distributed in the peripheral areas and some open areas, with leaf area densities mostly below 5 / m², presumably areas covered by low shrubs or sparse vegetation.

[0152] Spatial heterogeneity analysis: Over a horizontal range of approximately 0.36 miles (about 579 meters) covering the longitude range (129° 10' 30" E to 129° 11' 30" E), leaf area density varied significantly over short distances, with local density differences exceeding 40 / m², reflecting the complexity and diversity of the vegetation canopy density within the protected area. High- and low-value areas were interspersed, forming a distinct density gradient, which may have a significant impact on local light environments (such as light intensity) and water cycles (such as transpiration).

[0153] Ecological significance: Leaf area density data provides a basis for analyzing the relationship between vegetation structure and ecological function. High-density areas (>40 / m²) may become the core areas of forest productivity due to dense foliage and high photosynthetic efficiency, and are also major contributors to carbon sequestration. Low-density areas (<10 / m²) may have sparse foliage and poor light conditions, with low shrubs or herbaceous plants distributed there. In addition, the vertical stratification of leaf area density may affect forest light interception and energy distribution.

[0154] A composite strategy combining statistical outlier removal (SOR) and isolated point voxel filtering (IVF) is employed to achieve efficient denoising of raw point cloud data. 2) Progressive morphological filtering (PMF) and cloth simulation filtering (CSF) are used in synergistic processing to achieve accurate land cover classification and digital terrain model (DTM) construction for complex forest understory. 3) Based on Beer-Lambert physical laws and three-dimensional voxelization methods, the vertical distribution profile of the LAD (Laser Aperture Depression) is accurately inverted through quantitative analysis of laser pulse penetration within the canopy. This enables efficient, accurate, and reproducible estimation of the three-dimensional structural parameters of coniferous vegetation such as Korean pine, providing key technical support and data foundation for refined forest resource management, ecosystem function assessment, and regional carbon sequestration research. This addresses the technical bottlenecks of low efficiency and high cost in traditional LAD measurement methods, as well as the insufficient accuracy of existing remote sensing technologies in inverting the three-dimensional structure of complex coniferous forest canopies.

[0155] By employing a three-dimensional voxelization method, the macroscopic forest canopy is deconstructed into microscopic three-dimensional units, and the LAD (Lower Advection) is calculated within each unit. This not only yields an overall LAI (Lower Advection Index) value but also provides a detailed vertical distribution profile of the LAD. Figure 6 The figure shows a schematic diagram of the Leaf Area Density (LAI) profile calculated using the method proposed in this invention for calculating the leaf area density of red pine in practical application. As can be seen from the figure, the leaf area density profile is basically consistent with the vertical distribution of leaves in the region; the more lush the foliage within each horizontal layer, the larger the LAD of that layer. In the lower part of the canopy, LAD increases with height, reaching its maximum at a height of 20m; thereafter, the LAD value gradually decreases with increasing height. This profile reveals the vertical distribution pattern of photosynthetically active substances within the canopy, such as the main concentration of leaves within the canopy at certain height ranges. This refined three-dimensional structural information is unattainable by traditional two-dimensional remote sensing technology and has irreplaceable value for understanding deep-seated ecological issues such as the microclimate environment, light energy utilization efficiency, and species coexistence mechanisms within forests.

[0156] As can be seen, based on lidar data from the Wuying Nature Reserve in Yichun, Heilongjiang Province, digital terrain model (DTM), canopy height model (CHM), leaf area density (LAD), and leaf area index (LAI) were constructed, achieving efficient and accurate estimation of Korean pine vegetation structure parameters. The above results indicate that:

[0157] (1) Estimation of LAD and LAI: Using lidar point cloud data, combined with voxelization and Beer-Lambert's law, the LAD and LAI of Korean pine were successfully inverted. LAD reflects the leaf distribution characteristics of vegetation in the vertical direction, while LAI provides information on the total leaf area per unit surface area. Together, they provide key parameters for understanding the functions and processes of forest ecosystems.

[0158] (2) Technological advantages: With its high precision, high resolution, and all-weather operation, lidar technology can penetrate vegetation layers and obtain detailed three-dimensional structural information. This study demonstrates the effectiveness and superiority of lidar in the inversion of Korean pine vegetation structural parameters, especially under complex terrain and vegetation conditions.

[0159] (3) Ecological significance: The spatial distribution characteristics of LAD and LAI reveal the vertical structure and growth status of Korean pine vegetation. Areas with high LAD and LAI values ​​indicate dense vegetation growth and strong photosynthetic capacity, while areas with low values ​​may be sparsely vegetated or have low vegetation cover. This information is of great significance for forest management and protection.

[0160] This study not only improved the accuracy and efficiency of estimating Korean pine vegetation structure parameters, but also provided new technical means and theoretical foundations for research and applications in related fields. In the future, with the continuous development and application of lidar technology, its potential in forest ecosystem research will be further unleashed.

[0161] System implementation details:

[0162] On the other hand, this invention also proposes a system for calculating the area density of red pine needles. In practical applications, this system includes a UAV lidar data acquisition system, a data receiving system, and a data processing system. Its overall workflow is as follows:

[0163] 1. Data Acquisition: Using an UAV lidar data acquisition system integrating high-precision IMU / GNSS, high-density (e.g., > 50 points / m²) data of the Korean pine forest in the study area were obtained. 2 High-precision (e.g., elevation accuracy ±10 cm) 3D point cloud data.

[0164] 2. Data reception and processing: The collected raw data is transmitted to the ground data receiving system for trajectory processing and point cloud coordinate processing, generating raw point cloud data with accurate three-dimensional geographic coordinates.

[0165] 3. Data Processing: The solved point cloud data enters the core data processing system, where the data preprocessing module (noise removal, ground point classification, height normalization) and the model building and LAD calculation module (CHM generation, LAD inversion, LAI integration) are executed sequentially. Finally, high value-added products such as the vertical profile of leaf area density (LAD) and the spatial distribution map of leaf area index (LAI) of the target area are output.

[0166] Alternatively, the specific implementation of the system can also refer to the specific implementation methods 1-4 described above.

[0167] In summary, the composite preprocessing workflow proposed in this invention, through the combination of SOR and IVF, and the synergy of PMF and CSF, can systematically solve the challenges of noise in raw point clouds and feature classification under complex terrain. Compared to traditional workflows that rely on a single algorithm, the normalized point cloud data generated by this method is of higher quality, and the terrain model is more accurate. This solid data foundation ensures the accuracy of subsequent LAD and LAI inversion from the source and is the cornerstone of the overall system's high performance.

[0168] A physical model based on the Beer-Lambert law was used for LAD inversion, rather than a simple statistical regression model. The physical model has greater universality and interpretability because it is built upon the fundamental physical processes of laser-canopy interaction. This means that once the model parameters (primarily the extinction coefficient k) are properly calibrated for specific forest types and sensors, the model has the potential to be applied at different times and locations, yielding results that are more robust and reliable than empirical models that rely on specific training samples.

[0169] This invention addresses the core technical challenges of accurately estimating leaf area density (LAD) in Korean pine forests using UAV-based LiDAR technology. By constructing a comprehensive solution integrating composite data preprocessing, physical model inversion, and automated processes, this invention achieves high-precision, high-efficiency, and reproducible quantification of the three-dimensional structural parameters of complex coniferous forest canopies. Research results show that this method can effectively remove data noise, accurately separate ground features, and invert detailed LAD vertical profiles and LAI spatial distributions based on well-defined physical principles. This invention not only provides a powerful technical tool for scientific research and resource management of Korean pine forests but also lays a solid methodological foundation for extending LiDAR remote sensing technology to a wider range of forest ecosystem monitoring and assessment.

Claims

1. A method for calculating the area density of red pine needles, characterized in that, Includes the following steps: 1) Perform preprocessing operations on the original point cloud dataset of the red pine to be measured to obtain a point cloud dataset based on the ground surface; 2) Use the pulse-to-raster algorithm to convert the point cloud dataset based on the ground surface into a canopy height model that reflects the spatial height distribution of the red pine. 3) Based on the canopy height model, vertical stratification and voxel division are performed to obtain several height layers in the vertical direction and the number of data points and spatial density of each layer; according to the several height layers and the number of data points and spatial density of each layer, the corresponding LAD is calculated layer by layer using the Beer-Lamber law.

2. The method for calculating the area density of red pine needles according to claim 1, characterized in that, Based on several height layers and the number of data points and spatial density of each layer, the LAD of each layer is calculated using the Beer-Lambert law and the following formula: ； in, Let LAD be the value of the i-th layer; This represents the point cloud density at the top of the canopy (unobstructed area); Let be the spatial density of the i-th layer; k is an empirical coefficient.

3. The method for calculating the area density of red pine needles according to claim 1, characterized in that, Based on the canopy height model, vertical stratification and voxel partitioning were performed. The following method was used to obtain several height layers in the vertical direction, as well as the number of data points and spatial density of each layer: In three-dimensional space, the canopy height model is divided into several height layers according to a preset unit cube grid, and the height range of each height layer is calculated. Based on the height range of each height layer, the number of data points and spatial density of each height layer are calculated.

4. The method for calculating the area density of red pine needles according to claim 3, characterized in that, The height range of each height layer is expressed by the following formula: ； in, dz represents the height range of the i-th layer; dz represents the vertical resolution. This represents the minimum height range.

5. The method for calculating the area density of red pine needles according to claim 1, characterized in that, When calling the pulse-to-raster algorithm, each pulse in the point cloud dataset based on the ground surface is converted into a raster cell, and the height value of each raster cell is the highest elevation value of all point cloud data in that cell, thus generating the canopy height model of the red pine.

6. The method for calculating the area density of red pine needles according to claim 1, characterized in that, The preprocessing operation includes noise removal. During noise removal, a statistical discrete cluster removal algorithm and an isolated point voxel filtering algorithm are used to eliminate noise points in the original point cloud dataset. The statistical discrete cluster removal algorithm is used to remove outliers from all points in the original point cloud dataset. The isolated point voxel filtering algorithm is used to divide the point cloud space corresponding to the original point cloud dataset into a three-dimensional voxel grid of a preset fixed size; by counting the number of points in each voxel grid, all points in voxel grids with a number of points lower than a preset point number threshold are marked as noise points and removed.

7. The method for calculating the area density of red pine needles according to claim 6, characterized in that, The preprocessing operation also includes ground point classification. During ground point classification, a progressive morphological filtering algorithm and a cloth simulation filtering algorithm are used to separate the denoised original point cloud dataset into ground points and non-ground points. The progressive morphological filtering algorithm performs morphological opening operations on the denoised original point cloud dataset through a dynamically increasing window, thereby separating non-ground points from the denoised original point cloud dataset. The fabric simulation filtering algorithm is used to identify ground points in the denoised original point cloud dataset.

8. The method for calculating the area density of red pine needles according to claim 7, characterized in that, The preprocessing operation also includes point cloud elevation value normalization. During this normalization process, the relative ground height of vegetation is calculated using an irregular triangular mesh interpolation algorithm and a height normalization algorithm to obtain a point cloud dataset based on the land surface. Using the divided ground point dataset, a raster digital elevation model is generated through an irregular triangular mesh interpolation algorithm; The height normalization algorithm is used to subtract the elevation value in the digital elevation model of each point in the divided ground point dataset from the original elevation value of that point to obtain the relative ground height of that point; based on the relative ground heights of all points in the divided ground point dataset, the point cloud dataset with the ground surface as the reference is obtained.

9. The method for calculating the area density of red pine needles according to claim 1, characterized in that, Also includes: The leaf area index is calculated using the following formula: ； LAI stands for Leaf Area Index. Let LAD be the layer i; is the thickness of the i-th layer; n is the total number of layers.

10. A system for calculating the area density of red pine needles, characterized in that, Includes a processor for performing the method steps as described in any one of claims 1-9.

Images