A method and system for extracting three-dimensional structure parameters of an agroforestry shelterbelt

Abstract

The application discloses a kind of extraction method and system of farmland shelterbelt three-dimensional structure parameters, belong to remote sensing surveying and mapping and forestry information technical field.The method includes: obtaining the oblique photography point cloud data of target area;Through deep learning model, corresponding point cloud data of farmland shelterbelt is automatically identified and extracted;Based on the point cloud, closed three-dimensional model representing the overall outer contour of forest belt is constructed;Based on the model, the overall three-dimensional structure parameters of forest belt are automatically calculated, and the parameters include length, width, height, trend and the porosity of its ventilation and light transmission performance.The application innovatively proposes the through coefficient model fusing density difference, spatial coherence and feature connectivity length, and realizes the scientific quantification of porosity.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing mapping and forestry information technology, and particularly relates to a method and system for extracting three-dimensional structural parameters of farmland shelterbelts. Background Technology

[0002] Farmland shelterbelts, as an important component of the agricultural ecosystem, are typically planted in strips along farmland boundaries, roads, or irrigation canals. Multiple strips intertwine spatially to form a network structure, playing a core protective role in regulating the farmland microclimate, improving crop growth conditions, and ensuring stable and increased crop yields. They are irreplaceable in protecting the crop growth environment, enhancing agricultural production stability, and achieving coordination between farmland ecology and production functions. Compared to other shelterbelts, the core protective function of farmland shelterbelts is determined by their overall three-dimensional structure and spatial pattern, including length, width, and orientation. These parameters are the core basis for the planning, tending, regeneration, benefit evaluation, and optimization of agricultural production layout for farmland shelterbelts. Currently, the methods for investigating the three-dimensional structural parameters of shelterbelts are mainly divided into two categories: traditional ground measurement methods and extraction methods based on conventional remote sensing technology, both of which have significant limitations.

[0003] Traditional ground-based surveying methods rely on staff carrying total stations and surveying tools to conduct point-by-point measurements on-site. This method is entirely manual, with surveyors often taking temporary standard plots of 50-100m in length to measure the width and estimate the height of the shelterbelt, resulting in high labor intensity and low efficiency. Furthermore, manual surveys are highly subjective, have low accuracy, and rely heavily on experience, making it difficult to accurately determine the actual boundaries of shelterbelts on the ground in densely populated areas. This "small-scale, low-efficiency, low-precision" survey method is unsuitable for the "large-scale, network-like distribution" of farmland shelterbelts, and cannot meet the demands of modern forestry for precise data. Extraction methods based on conventional remote sensing technologies mainly include satellite remote sensing and UAV orthophoto technology. While satellite remote sensing technologies such as the Gaofen series satellite data can achieve macroscopic monitoring of large-scale network-like shelterbelts, their low spatial resolution makes it difficult to distinguish the boundaries between the strips and surrounding farmland, resulting in significant errors in measuring parameters such as shelterbelt width and orientation angles. Although the resolution of UAV orthophoto technology has improved, it can only acquire two-dimensional information and cannot directly obtain core parameters such as canopy height and three-dimensional volume. It needs to be combined with ground sample data for indirect calculation, which is a complex process and the accuracy is limited by the representativeness of the sample points, making it difficult to support the precise management of shelterbelts.

[0004] While LiDAR technology is used for high-precision extraction of forestry parameters, it is mismatched with the needs of farmland shelterbelt surveys. Specifically: 1) High cost and difficulty in promotion: professional airborne equipment costs 500,000 to 2 million yuan, and the rental cost per square kilometer is tens of thousands of yuan, making it unsuitable for large-scale surveys; 2) Data redundancy and low efficiency: it can obtain point cloud densities of 500-1000 points / square meter, including details such as branch distribution that are not needed in shelterbelt surveys, while key parameters such as length and width require time-consuming secondary processing to obtain. The complete process per square kilometer takes 2-3 days, which cannot meet the timeliness requirements of farmland shelterbelt tending and assessment; 3) High barrier to entry and strong professionalism: data quality depends on specialized technical personnel, and the equipment such as professional graphics workstations required for massive data processing is difficult for grassroots departments to provide. Compared with LiDAR, oblique photogrammetry has relatively lower usage costs and higher survey efficiency. However, when this method is currently applied in forestry surveys, it only converts the three-dimensional modeling information into DSM, i.e., digital surface model, for reuse and analysis. This greatly reduces the amount of data information, details and accuracy obtained by oblique photogrammetry, and makes it impossible to obtain accurate forest belt structure parameters.

[0005] In summary, existing methods either focus on individual tree details while neglecting the overall three-dimensional features of the forest belt, or while covering a large area, lack sufficient accuracy, or achieve the required accuracy but suffer from a cost-efficiency imbalance. The core problem lies in failing to address the fundamental requirement of forest belt surveys regarding the overall three-dimensional structure of the forest belt, and lacking a highly efficient technical system adapted to farmland scenarios and aimed at obtaining three-dimensional structural parameters of the forest belt. Therefore, this invention proposes a rapid extraction method for three-dimensional structural parameters of forest belts based on oblique photogrammetry, fulfilling the practical needs of efficient, accurate, low-cost, and large-area forest belt structure surveys. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method and system for extracting three-dimensional structural parameters of farmland shelterbelts.

[0007] The specific details of the invention are as follows:

[0008] A method for extracting three-dimensional structural parameters of farmland shelterbelts includes the following steps: acquiring oblique photogrammetric point cloud data of a target area, wherein the target area contains continuously distributed strip-shaped forest belts; automatically identifying and extracting forest belt point cloud data corresponding to the farmland shelterbelt from the oblique photogrammetric point cloud data; constructing a closed three-dimensional model representing the overall outer contour of the farmland shelterbelt based on the forest belt point cloud data; and automatically calculating at least one overall three-dimensional structural parameter of the farmland shelterbelt based on the closed three-dimensional model, wherein the overall three-dimensional structural parameter includes at least one of length, width, height, orientation, and permeability representing its overall shape and spatial distribution.

[0009] Preferably, the overall three-dimensional structural parameters also include permeability, which characterizes the ventilation or light transmission performance of the farmland shelterbelt.

[0010] Preferably, the automated calculation of at least one overall three-dimensional structural parameter of the farmland shelterbelt includes:

[0011] Determine the geometric center axis of the closed three-dimensional model;

[0012] Based on the geometric center axis, calculate the length of the farmland shelterbelt, the width perpendicular to the axis, and the direction angle of the axis.

[0013] Preferably, the automated calculation of the sparseness of the farmland shelterbelt includes:

[0014] Based on the closed three-dimensional model or the point cloud data of the forest belt, a virtual probe is simulated to be launched into the interior of the farmland shelterbelt.

[0015] The permeability is calculated based on the interaction information between the virtual detector and the farmland shelterbelt.

[0016] Preferably, the simulation of emitting a virtual detector into the farmland shelterbelt specifically involves simulating multiple virtual rays parallel to a preset wind direction passing through the cross-section of the farmland shelterbelt.

[0017] The permeability is calculated specifically by: calculating the proportion of the multiple virtual rays that do not interact with the forest belt, or the average path length of the virtual rays that interact with the forest belt.

[0018] Preferably, the calculation of the permeability specifically involves calculating the connectivity coefficient of the farmland shelterbelt. The connectivity coefficient is a dimensionless parameter obtained by combining density difference, spatial autocorrelation, and characteristic connectivity length. The density difference is used to characterize the degree of difference in the density distribution of the farmland shelterbelt in the cross section, the spatial autocorrelation is used to characterize the spatial coherence of the pore unit distribution in a preset direction, and the characteristic connectivity length is used to characterize the average scale of the coherent channels formed by the pore units in the preset direction.

[0019] Preferably, the automatic identification and extraction of point cloud data of the forest belt corresponding to the farmland shelterbelt includes:

[0020] Based on the continuous spatial distribution morphology of the farmland shelterbelt in point cloud data, the farmland shelterbelt is classified and extracted from background features.

[0021] A system for extracting three-dimensional structural parameters of farmland shelterbelts, comprising:

[0022] The data acquisition module is used to acquire oblique photogrammetric point cloud data of the target area, which includes forest belts that are continuously distributed in strips.

[0023] The point cloud processing module is used to automatically identify and extract the point cloud data of the forest belt corresponding to the farmland shelterbelt from the oblique photogrammetry point cloud data;

[0024] The 3D modeling module is used to construct a closed 3D model representing the overall outer contour of the farmland shelterbelt based on the forest belt point cloud data;

[0025] The parameter calculation module is used to automatically calculate at least one overall three-dimensional structural parameter of the farmland shelterbelt based on the closed three-dimensional model. The overall three-dimensional structural parameter includes at least one of length, width, height, orientation, and permeability that characterize its overall shape and spatial distribution.

[0026] A computing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for extracting three-dimensional structural parameters of farmland shelterbelts as described above.

[0027] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a method for extracting three-dimensional structural parameters of farmland shelterbelts as described above.

[0028] The benefits of this invention are as follows: by integrating oblique photography, deep learning, and an innovative connectivity coefficient model, it achieves integrated, automated, and rapid extraction of three-dimensional structural parameters of farmland shelterbelts. While ensuring measurement accuracy, it increases survey efficiency by tens of times, significantly reduces equipment and labor costs, and for the first time realizes the scientific quantification and direct three-dimensional calculation of the key indicator of forest belt ecological function—permeability. This solves the problems of low efficiency, high cost, and lack of functional evaluation in existing technologies, providing an efficient and reliable technical means for precise monitoring and management of forestry resources. Attached Figure Description

[0029] Figure 1 :flow chart;

[0030] Figure 2 Data collected by UAV: ​​a) Oblique photogrammetry 3D reconstruction model, b) LAS point cloud data;

[0031] Figure 3 : Schematic diagram of forest belt point cloud classification results, a) Classified results, b) Forest belt LAS point cloud data exported separately;

[0032] Figure 4: Schematic diagram of the construction process of the three-dimensional model of the forest belt, a) Three-dimensional reconstruction model of a single forest belt in an instance plot, b) Point cloud after preprocessing, c) TIN polyhedron model, d) Simplified closed cube model;

[0033] Figure 5 : Schematic diagram of the principle of extracting three-dimensional structural parameters of forest belt, a) closed cube model in vertical coordinate system, b) central axis and parameters of cube. Detailed Implementation

[0034] The following embodiments illustrate the present invention in detail. All raw materials and equipment used in the present invention are commercially available products and can be directly obtained through market purchase.

[0035] The present application will be further described in detail below with reference to embodiments, comparative examples and performance test results. These embodiments should not be construed as limiting the scope of protection claimed in this application.

[0036] In the following description of the embodiments, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0037] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0038] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0039] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0040] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0041] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0042] Example 1: Technical Description

[0043] This embodiment is used to explain in detail the method and how the system is specifically implemented. The core process of this invention follows four main steps, and combined with preferred technical solutions, ultimately outputting the overall three-dimensional structural parameters of the forest belt, including innovative "permeability" and, in particular, "connectivity coefficient".

[0044] 1. Data Acquisition

[0045] In practice, a drone equipped with a five-lens oblique photography camera, such as the DJI M300RTK, is used to plan flight routes for the area containing the target forest belt. The flight altitude is set at 80-120 meters, with both the heading and lateral overlap rates not less than 70% to ensure the integrity of the multi-view images. The acquired images are processed using commercial software, such as ContextCapture or DJI Terra, for aerial triangulation and density matching, ultimately generating LAS format point cloud data containing three-dimensional coordinates (X, Y, Z) and color information. This point cloud data density typically reaches over 100 points per square meter, sufficient to clearly depict the three-dimensional morphology of the forest belt and surrounding terrain features.

[0046] 2. Forest Belt Point Cloud Identification and Extraction

[0047] The implementation includes two phases:

[0048] Point Cloud Classification: The acquired overall point cloud is imported into a processing platform such as ArcGIS Pro or CloudCompare. A deep learning-based point cloud semantic segmentation algorithm is used to automatically classify the point cloud. When training the network, for the "forest belt" category, the significant strip-like and continuous spatial morphological features it exhibits in the point cloud are emphasized as one of the core training labels, enabling it to be effectively distinguished from scattered trees, buildings, roads, and other features. After classification, the forest belt point cloud is assigned a specific category label.

[0049] Point cloud extraction: Based on the classification labels, all point cloud data belonging to the "forest belt-forest belt" category are extracted from the original data in batches to generate a clean 3D point cloud dataset containing only the target forest belt, laying the foundation for subsequent modeling.

[0050] 3. Three-dimensional model construction

[0051] Point cloud preprocessing: The extracted forest belt point cloud is thinned and denoised to reduce the amount of data and noise while preserving the overall shape.

[0052] Surface modeling: The preprocessed point cloud is used to generate a TIN model using an irregular triangular mesh construction algorithm. This model consists of a large number of triangular facets that precisely fit the undulations of the forest canopy surface.

[0053] Model Closure: Based on the "top" and "side" surfaces of the forest belt represented by the TIN model above, and combined with the ground elevation at the bottom of the forest belt extracted from the point cloud or obtained through interpolation, a "bottom" surface is added to the model, thus forming a watertight, polyhedral 3D model with a closed bottom. This model no longer distinguishes individual trees, but abstracts and represents the 3D spatial extent occupied by the entire forest belt, which is the basis for subsequent calculations of volume and structural permeability.

[0054] 4. Automated extraction of three-dimensional structural parameters

[0055] 4.1 Extraction of basic geometric parameters

[0056] Based on the aforementioned closed 3D model, the system automatically executes:

[0057] Determine the geometric center axis: Calculate the three-dimensional geometric center of the model and fit a spatial curve along its maximum extension direction as the central axis of the forest belt.

[0058] Calculation parameters:

[0059] Length (l): The length of the spatial curve of the central axis.

[0060] Width (d): The maximum lateral span of the model is measured on multiple sections perpendicular to the central axis, and the average value is taken.

[0061] Height (h): The elevation difference between the top of the model and the closed bottom surface is calculated at multiple representative locations, and the average value is taken.

[0062] Orientation (A): Project the central axis onto the horizontal plane and calculate the angle between the projected line and the due north direction.

[0063] 4.2 Extraction of Permeability Parameters – Two Specific Implementation Methods

[0064] Implementation Method A: Statistical Method Based on Virtual Ray Penetration

[0065] The physical meaning of this method is intuitive. The system simulates the emission of a large number of parallel and uniformly distributed virtual rays along a predetermined dominant direction, guiding them through a three-dimensional model of the forest belt. Subsequently, any of the following indicators is statistically analyzed and calculated as permeability:

[0066] 1. X-ray penetration rate

[0067] in, The number of rays that pass completely through the model without intersecting the model surface. This represents the total number of rays.

[0068] 2. Average path length (L):

[0069] in, Let be the length of the i-th ray traveling inside the model. This represents the total number of rays intersecting with the model. The smaller the value, the easier it is for rays to penetrate, and the higher the permeability.

[0070] Implementation Method B: Connectivity Coefficient Model Based on Spatial Coherence

[0071] This method proposes a novel penetration coefficient. The mathematical model, which comprehensively considers density differences, spatial correlations, and connectivity scales, can more scientifically assess the ventilation and light transmission potential of forest belts. The calculation formula is as follows:

[0072] in: Density variation coefficient: reflects the degree of undulation in the density distribution within the forest belt. The calculation method involves discretizing the forest belt model on a cross-section and calculating the density of each unit. Standard deviation with the mean The ratio: Spatial autocorrelation: quantifies the correlation of pore spatial distribution in a specific direction. It is calculated based on the autocorrelation function in spatial statistics. First, a pore indicator function is defined. If the value is 1 at pores and 0 otherwise, the autocorrelation value at a specific lag distance d is:

[0073] Take a feature distance For example, average crown width, . The higher the value, the stronger the coherence of the pore arrangement. The scale decay function is used to evaluate the effectiveness of coherent pore channels. This represents the average characteristic length of the pore connectivity along the analytical direction. The function form is as follows:

[0074] Where λ is the reference length (e.g., the average width of the forest belt). This function ensures that the value is only determined when the length of the continuous pores is... Only when it is large enough (comparable to λ) does it make a substantial contribution to overall connectivity.

[0075] Calculation example: Suppose the calculation for a certain forest belt model yields... ,but:

[0076] This value quantitatively reflects the overall connectivity performance of the forest belt.

[0077] 5. System Implementation

[0078] This system can be built as a software platform or integration tool that includes the following logical modules:

[0079] Data acquisition module: responsible for accessing raw images or pre-processed point cloud data collected by drones.

[0080] Point cloud processing module: It has a built-in deep learning classification model to perform automatic identification, classification and extraction of forest belt point clouds.

[0081] 3D modeling module: integrates point cloud thinning, filtering, TIN construction, and model closure algorithms.

[0082] Parameter calculation module: This is the core module, integrating algorithms such as geometric parameter calculation, ray penetration statistics, and penetration coefficient (TCC) model, which can automatically output a complete three-dimensional structural parameter report of the forest belt.

[0083] Example 2: Standard Implementation Process

[0084] Subject and Objective: The research object is farmland shelterbelt in Changtu County, Tieling City, Liaoning Province. The objective is to implement the entire process from data collection to parameter extraction on a representative section of the shelterbelt to verify the feasibility of the method and the reliability of the results.

[0085] Implementation process and data:

[0086] Data collection:

[0087] A DJI M3M drone equipped with an oblique photography module was used. The planned flight altitude was 100 meters, with both the heading and lateral overlap rates set at 80%. The tilt angles of the oblique camera were -30°, -45°, and -60°. Oblique images covering approximately 0.4 km² were acquired.

[0088] The images were processed using DJI Terra V4.5.6 software, with distortion correction enabled, high-precision aerial triangulation mode employed, and three ground control points measured by RTK added. The final result was a LAS 1.4 format point cloud with a density of approximately 100 points per square meter. Figure 2 (a) shows the 3D reconstruction model generated after data processing. Figure 2 (b) shows the corresponding LAS point cloud data, with the sample plot area within the red box.

[0089] Forest belt point cloud classification and export:

[0090] Import the point cloud into ArcGIS Pro and perform statistical filtering to remove noise. Then, use the "Detect Objects from Point Clouds Using Trained Models" tool for deep learning classification.

[0091] Specific operations: After slicing the point cloud, 80% is used for training and 20% for validation. The "point cloud object detection" architecture is selected, with the "high forest belt" class code (5) as the target, and the model is trained until the accuracy is ≥95%. The model is applied to classify the point cloud of the whole region, and the forest belt point cloud is automatically labeled and extracted. Figure 3 (a) shows the point cloud results after classification (different colors represent different land cover categories). Figure 3 (b) is the derived point cloud of the pure forest belt.

[0092] 3D model construction:

[0093] Point cloud optimization: The exported forest belt point cloud is thinned using a 0.5m × 0.5m grid and then statistically filtered again to obtain an optimized point cloud, such as... Figure 4 (b)

[0094] TIN Model Generation: In ArcGIS Pro's 3DAnalyst module, use the "Create TIN" tool, with the optimized point cloud as "hard break line" input, setting the maximum side length to 2m and the Z tolerance to 0.3m, to generate an irregular triangular mesh (TIN) model, such as... Figure 4 As shown in (c).

[0095] Model simplification and closure: The model was simplified using the "Simplify TIN" tool (Z tolerance 0.3m), and then the "Construct LAS Polyhedron" tool was used to generate a three-dimensional polyhedron model with a closed bottom, using the measured ground elevation (87.8m) as the base and combining it with the forest belt boundary polygons. Figure 4The simplified closed model shown in (d)

[0096] Extraction of three-dimensional structural parameters of forest belts:

[0097] In ArcGIS Pro, the geometric center (axis) is automatically calculated using the 3DAnalyst tool based on a closed 3D model.

[0098] Parameters are automatically extracted based on the axis and model geometry. Figure 5 The principle of parameter extraction is intuitively demonstrated: Figure 5 (a) is a schematic diagram of the closed cube model of the forest belt in the vertical coordinate system. Figure 5 (b) specifies the measurement methods for the centerline length (l), vertical axis width (d), height (h), and centerline direction angle (A).

[0099] Repeated measurements were performed on the sample plots, and the results are shown in Table 1. Taking sample point 1 as an example, the average length of the forest belt was 500.8 meters, the average width was 20.2 meters, the average height was 18.3 meters, and the orientation was 105.5°. The average length of the sample plot was 469.2 meters, the average width was 20.2 meters, and the average height was 18.5 meters. This embodiment... Figure 2 , Figure 3 , Figure 4 , Figure 5 The series of presentations in Table 1 fully and meticulously reproduces the standard implementation process.

[0100] Example 3: Adaptability Verification in Complex Terrain

[0101] The invention was implemented in complex hilly terrain to verify its terrain adaptability. The implementation target was a windbreak and sand-fixing forest belt planted along contour lines in a mountainous area, with a terrain elevation difference of approximately 80 meters.

[0102] Implementation process and data:

[0103] 1. Data Acquisition: A DJI M300RTK drone was used, equipped with a Zenmuse P1 full-frame oblique photography camera. A terrain-following flight path was planned, with a relative flight altitude of 80 meters, a directional overlap of 85%, and a lateral overlap of 75%. The flight operation acquired 2874 raw images. Figure 2 As shown in (a), a three-dimensional reconstruction model of the area after processing is displayed, and the distribution of the forest belt along the terrain undulations can be clearly seen.

[0104] 2. Point Cloud Generation and Terrain Processing: Aerial triangulation and dense reconstruction were performed using ContextCapture software to generate LAS point clouds (point density approximately 90 points / square meter). Due to the complex terrain, the elevation values ​​of forest belts, bare slopes, and terraced ridges in the point cloud highly overlapped. To address this issue, a high-precision digital terrain model (DTM) was first extracted from the original point cloud using the Progressive Triangulation Decryption Filter (PTD) algorithm, such as... Figure 4 (b) shows a portion of the area, providing an accurate ground reference for subsequent separation of forest belts.

[0105] 3. Intelligent Extraction of Forest Belt Point Clouds: This method employs deep learning classification that combines morphological and geographical context. The specific process involves training a PointNet++ network model using the original point cloud (containing XYZ and RGB information) and the extracted DTM elevation difference (nDSM) as input features. This network specifically learns the composite features of "continuous distribution along contour lines" and "above a certain terrain threshold." After classification, the forest belt point clouds are successfully extracted, as shown below. Figure 3 As shown in (a), different colors in the figure represent different land cover categories (green for forest belts, gray for ground, blue for buildings, etc.), and the forest belts are clearly separated against a complex background. Figure 3 (b) shows the forest belt point cloud exported separately from the classification results.

[0106] 4. 3D Model Construction and Terrain Correction: The extracted forest belt point cloud was first thinned using a 0.5m grid, and then outliers were removed using statistical filtering (20 neighborhood points, standard deviation multiple of 2.0). Then, in ArcGIS Pro, the "Create TIN" tool was used, with the point cloud as the "hard break line" input and the maximum side length set to 3m, to construct a preliminary TIN polyhedron model, the shape of which is as follows... Figure 4 As shown in (c), the triangular mesh closely fits the undulating canopy surface. Subsequently, the bottom of the TIN model is clipped to the DTM surface, generating a closed 3D model that perfectly conforms to the terrain (i.e., the bottom of the model is the terrain surface, and the top is the canopy surface), as shown in [example image]. Figure 4 As shown in (d), this is a simplified closed model diagram.

[0107] 5. Parameter Extraction and Results: The system automatically calculates model parameters. Taking one forest belt as an example, the extraction results are as follows: the length of the central axis (along the terrain surface) is 458.3 meters, the average width perpendicular to the central axis is 28.7 meters, and the forest belt orientation (projected onto the horizontal plane) is 152.4°. Most importantly, the forest belt height is calculated from the average elevation difference between the top curved surface and the bottom DTM curved surface of the model, which is 16.8 meters. This accurately reflects the net growth height of the forest belt relative to its site, rather than its absolute altitude. This embodiment uses... Figure 2 (a) Figure 3 (a)(b) Figure 4The series of illustrations and data processing flowcharts in (b)(c)(d) fully demonstrate the entire process of this invention from data to model to parameters in complex terrain.

[0108] Example 4: Calculation and Functional Verification of Permeability Parameters

[0109] Implementation process and data:

[0110] 1. Basic 3D Model: A high-precision closed 3D model of the forest belt was constructed using the method described in Example 1. The basic geometric parameters of the model are: length 600.5 meters, average width 24.6 meters, and average height 24.6 meters (see Table 1).

[0111] Table 1. Measurement results of three-dimensional structural parameters of forest belts in the sample plots.

[0112] 2. Virtual ray method calculation:

[0113] Principles and settings: such as Figure 5 (b) As shown in the schematic diagram, to simulate the wind penetration process, a set of parallel virtual rays are generated on the cross-section of the forest belt model (perpendicular to the forest belt direction at 110.7°). In this embodiment, the ray spacing is set to 0.05 meters, and a total of [number missing] virtual rays are generated. =width / 0.05≈492 rays.

[0114] Calculations and Results: The system simulates the intersection calculation between each ray and the 3D model. Statistical analysis shows that N_through = 118 rays completely pass through the model without intersecting any triangular faces. Ray penetration rate is calculated. Simultaneously, the average internal path length of all rays intersecting the model is calculated. =8.7 meters.

[0115] 3. Calculation using the continuity coefficient model method:

[0116] Parametric analysis: First, the forest belt model is divided into continuous cross-sectional layers along the wind direction (perpendicular to the direction). For each layer, it is discretized into 0.2m × 0.2m grid cells, and the density of each cell is calculated based on the number of point clouds to which it belongs. .

[0117] calculate Calculate the mean density of all cells. and standard deviation Then the density difference coefficient .

[0118] Calculate C: Set density threshold The unit is marked as a pore. The Moran index I, with a lag distance of d = 5m (approximately crown width), was calculated as the spatial autocorrelation C, and the calculated C = 0.58.

[0119] calculate and Connectivity analysis was performed on the pore elements in the 3D model to identify the main pore channels and calculate their projected lengths in the wind direction. The characteristic connectivity length was obtained by weighted averaging. =10.4 meters. (Use reference length) =24.6 meters (average width of the forest belt), calculate the scale function: .

[0120] Calculate TCC: Combining the above parameters, .

[0121] 4. Results Comparison and Discussion: The penetration rate P obtained by the X-ray method was 0.240, while the penetration coefficient TCC was 0.133. Figure 5 (b) vividly illustrates the physical reason for this difference: the ray method only counts the proportion of voids on the "line," while the TCC model... Further assessments were conducted on the distribution quality, coherence, and effective scale of these cavities on both the "surface" and "volume" surfaces. The lower TCC value indicates that although the forest belt has a certain direct permeability, the organization of these pore spaces does not form highly coherent ventilation corridors, thus the overall ventilation efficiency is lower than that predicted by simple porosity.

[0122] Example 5: Application of Automated Survey of Networked Forest Belt Systems

[0123] The demonstration area is a farmland shelterbelt with a total area of ​​about 15 square kilometers, which consists of dozens of crisscrossing forest belts.

[0124] Implementation process and data:

[0125] 1. Systematic data acquisition and processing:

[0126] Two unmanned aerial vehicle (UAV) systems were used in parallel to complete the oblique photogrammetry data acquisition for the entire area within 1.5 days. The acquired raw image data, exceeding 2TB, was imported into a high-performance graphics workstation cluster deployed with the "Farmland Shelterbelt 3D Structure Parameter Extraction System" described in this invention.

[0127] The system initiates a fully automated processing pipeline. First, it uses the standard interface of its data acquisition module to read images. Then, the aerial triangulation and dense reconstruction submodules within its point cloud processing module automatically run, generating a unified 3D reality model and LAS point cloud data for the entire area within 48 hours (with similar local effects). Figure 2 (a) and Figure 2 (b)).

[0128] 2. Batch intelligent information extraction:

[0129] The deep learning classification submodule in the system's point cloud processing module loads a pre-trained dedicated model to perform automated semantic segmentation of the entire point cloud area. For example... Figure 3 As shown in (a) (only a partial illustration), the system successfully separated all forest belts in the entire forest network from farmland, roads, and villages at once, and automatically assigned a unique ID to each forest belt.

[0130] The 3D modeling module then processes the point cloud of each forest belt in batches according to its ID, sequentially performing steps such as thinning, filtering, TIN construction, and model closure to generate an independent 3D model for each forest belt. The principle behind this construction process is similar to... Figure 4 It is completely consistent with what is shown.

[0131] 3. Fully automated calculation and output of parameters:

[0132] The system's parameter calculation module traverses more than 60 generated 3D models of forest belts. For each model, the module automatically performs the following calculations: calculating length, width, and orientation based on the geometric central axis; calculating basic permeability based on the virtual ray method; and calculating TCC based on the penetration coefficient model.

[0133] All calculation tasks are completed within 8 hours, and the system automatically summarizes and generates a structured database (such as CSV format) and thematic maps containing all forest belt parameters. Figure 1 The flowchart accurately summarizes the entire automated business chain from "data acquisition" to "parameter extraction".

[0134] 4. Application Results: The final output includes: 1) A 3D forest network scene (similar to...) Figure 2 a) 2) Vector distribution map of forest belts; 3) Parameter attribute table of forest belts. Managers can directly query detailed parameters of any forest belt through the GIS platform. For example, they can quickly filter out forest belt segments with TCC values ​​below 0.15 and potentially insufficient protective effectiveness, providing direct basis for precise tending decisions. This embodiment references... Figure 1 , Figure 2 , Figure 3 , Figure 4 This fully demonstrates how the system described in this invention transforms massive amounts of raw data into structured information that can be directly used for management decisions during a regional census, showcasing the significant advantages and application value of the system and equipment in achieving large-scale, automated, and intelligent surveys.

[0135] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for extracting three-dimensional structural parameters of farmland shelterbelts, characterized in that, Includes the following steps: Obtain oblique photographic point cloud data of the target area, which includes forest belts that are continuously distributed in strips; Automatically identify and extract the forest belt point cloud data corresponding to the farmland shelterbelt from the oblique photogrammetry point cloud data; Based on the forest belt point cloud data, a closed three-dimensional model representing the overall outer contour of the farmland shelterbelt is constructed. The closed three-dimensional model is obtained by generating an irregular triangular mesh model based on the forest belt point cloud data, and adding a bottom surface to the model to form a bottom-closed watertight polyhedron three-dimensional model. Based on the closed three-dimensional model, at least one overall three-dimensional structural parameter of the farmland shelterbelt is automatically calculated, and the overall three-dimensional structural parameter includes the permeability that characterizes its ventilation or light transmission performance. The permeability is calculated by: calculating the connectivity coefficient of the farmland shelterbelt, wherein the connectivity coefficient is a dimensionless parameter obtained by combining density difference, spatial autocorrelation, and characteristic connectivity length; the density difference is used to characterize the degree of difference in density distribution of the farmland shelterbelt in the cross section, the spatial autocorrelation is used to characterize the spatial coherence of the distribution of pore units in a preset direction, and the characteristic connectivity length is used to characterize the average scale of the coherent channels formed by the pore units in the preset direction.

2. The method for extracting three-dimensional structural parameters of farmland shelterbelts according to claim 1, characterized in that, The farmland shelterbelt also includes at least one overall three-dimensional structural parameter: The length, width, height, and orientation that characterize its overall shape and spatial distribution; Determine the geometric center axis of the closed three-dimensional model; Based on the geometric center axis, calculate the length of the farmland shelterbelt, the width perpendicular to the axis, and the direction angle of the axis.

3. The method for extracting three-dimensional structural parameters of farmland shelterbelts according to claim 1, characterized in that, The calculation of the connectivity coefficient of the farmland shelterbelt is specifically as follows: The continuity coefficient is calculated using the following formula. : in, For density difference, For spatial autocorrelation, To use the characteristic connectivity length Let be the scaling decay function of the variable; The density difference , The standard deviation of the density of each unit within the cross-section of the forest belt. This represents the average density of each unit cell. The spatial autocorrelation This is the autocorrelation value calculated based on the Moran index; The scale decay function , This represents the average width of the forest belt.

4. The method for extracting three-dimensional structural parameters of farmland shelterbelts according to claim 1, characterized in that, The automatic identification and extraction of point cloud data of the forest belt corresponding to the farmland shelterbelt includes: Based on the continuous spatial distribution morphology of the farmland shelterbelt in point cloud data, the farmland shelterbelt is classified and extracted from background features.

5. A system for extracting three-dimensional structural parameters of farmland shelterbelts, characterized in that, include: The data acquisition module is used to acquire oblique photogrammetric point cloud data of the target area, which includes forest belts that are continuously distributed in strips. The point cloud processing module is used to automatically identify and extract the point cloud data of the forest belt corresponding to the farmland shelterbelt from the oblique photogrammetry point cloud data; The 3D modeling module is used to construct a closed 3D model representing the overall outer contour of the farmland shelterbelt based on the forest belt point cloud data; wherein, the closed 3D model is obtained by generating an irregular triangular mesh model based on the forest belt point cloud data, and adding a bottom surface to the model to form a bottom-closed watertight polyhedron 3D model. The parameter calculation module is used to automatically calculate at least one overall three-dimensional structural parameter of the farmland shelterbelt based on the closed three-dimensional model. The overall three-dimensional structural parameter includes permeability, which characterizes its ventilation or light transmission performance. The permeability is calculated by calculating the connectivity coefficient of the farmland shelterbelt, which is a dimensionless parameter obtained by combining density difference, spatial autocorrelation, and characteristic connectivity length. The density difference is used to characterize the degree of difference in the density distribution of the farmland shelterbelt in the cross section. The spatial autocorrelation is used to characterize the spatial coherence of the pore unit distribution in a preset direction. The characteristic connectivity length is used to characterize the average scale of the coherent channels formed by the pore units in the preset direction.

6. The system for extracting three-dimensional structural parameters of farmland shelterbelts according to claim 5, characterized in that, The parameter calculation module automatically calculates the overall three-dimensional structural parameters, including: length, width, height, and orientation, which characterize its overall shape and spatial distribution.

7. A computing device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a method for extracting three-dimensional structural parameters of farmland shelterbelts as described in any one of claims 1 to 4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a method for extracting three-dimensional structural parameters of farmland shelterbelts as described in any one of claims 1 to 4.

Images