Forest cultivation dynamic monitoring method and system based on remote sensing technology

Abstract

The invention relates to a forest cultivation dynamic monitoring method and system based on a remote sensing technology, and the method comprises the steps: collecting point cloud and multi-source topographic data of a complex topographic region through the fusion of an airborne laser radar and a satellite radar technology, and generating a high-precision point cloud and topographic parameter matrix through denoising and registration; a segmentation threshold value is dynamically optimized in combination with topographic features, the canopy point cloud penetration rate of abrupt slope and valley areas is improved, and the problem of canopy missegmentation caused by uneven point cloud density under a complex terrain is solved; constructing a penetration rate compensation model driven by terrain influence factors, correcting laser radar point cloud deviation and inverting high-precision tree height, crown breadth and biomass parameters; and based on fusion analysis of the multi-temporal vegetation index and a machine learning model, early warning of diseases and insect pests and fire hazards and dynamic evaluation of a man-made forest cultivation effect are realized. According to the method, the bottleneck of complex terrain monitoring is broken through, the forest parameter inversion precision and the real-time response capability are remarkably improved, and technical support is provided for precise management of forestry resources.

Description

Technical Field

[0001] The present invention relates to the technical field of forest cultivation, and in particular to a forest cultivation dynamic monitoring method and system based on remote sensing technology. Background Art

[0002] Dynamic monitoring of forest cultivation is a core component of forestry resource management. It aims to provide a scientific basis for forestry planning and disaster prevention by continuously tracking tree growth, health, and environmental responses. Traditional monitoring methods rely primarily on manual plot surveys and remote sensing image interpretation. However, in complex terrain areas, high-precision, fully automated dynamic tracking is difficult due to factors such as dense vegetation cover and dramatically undulating terrain. In recent years, the integrated application of multi-platform remote sensing technologies, combined with three-dimensional LiDAR perception and satellite multispectral analysis, has gradually become a key direction for improving the accuracy of forest parameter inversion.

[0003] However, existing technologies still have significant limitations when dealing with complex terrains such as mountains and hills. Due to the undulating terrain, the penetration rate of LiDAR point clouds is unevenly distributed, the point cloud density in areas such as steep slopes and valleys drops sharply, and the problem of canopy structure missegmentation is prominent, which directly affects the inversion accuracy of key parameters such as tree height and biomass. At the same time, traditional methods mostly rely on static threshold segmentation and fail to dynamically optimize the point cloud processing process in combination with terrain characteristics, resulting in insufficient reliability of monitoring results in complex scenarios. In addition, the real-time performance of multi-temporal data fusion and anomaly detection is insufficient, making it difficult to meet the needs of rapid response to forest fires and outbreaks of pests and diseases. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide a forest cultivation dynamic monitoring method and system based on remote sensing technology that can adapt to complex terrain, improve the accuracy of penetration correction and support dynamic tracking.

[0005] The purpose of the present invention is achieved by the following scheme:

[0006] In a first aspect, the present invention provides a method for dynamic monitoring of forest cultivation based on remote sensing technology, comprising the following steps:

[0007] S1: Based on airborne LiDAR technology and satellite radar technology, the lidar point cloud data and multi-source terrain data in complex terrain areas are collected and preprocessed to generate high-precision registered point clouds and terrain parameter matrices after preprocessing;

[0008] S2: Based on the terrain parameter matrix, the high-precision registered point cloud is subjected to terrain adaptive segmentation. The canopy point cloud penetration rate is optimized by combining the dynamic threshold segmentation algorithm to obtain the canopy point cloud segmentation results of different terrain units. The canopy point cloud segmentation results are used to indicate the canopy cover density and individual tree parameters.

[0009] S3: Based on the canopy point cloud segmentation results and the terrain parameter matrix, a canopy penetration compensation model is constructed. The lidar point cloud data is corrected based on the canopy penetration compensation model to generate corrected forest 3D structural parameters, including tree height, canopy width, and biomass.

[0010] S4: Based on multi-temporal remote sensing data, the corrected three-dimensional forest structural parameters are processed, and the forest dynamic growth indicators are inverted through the normalized vegetation index time series and machine learning models to generate cultivation monitoring results. The cultivation monitoring results are used to indicate the risks of pests and diseases, fire hazards and the cultivation effects of artificial forests.

[0011] In one embodiment, the present invention provides a method for dynamic monitoring of forest cultivation based on remote sensing technology, step S1 specifically comprising the following steps:

[0012] S11: Perform multiple scans of complex terrain areas using UAV-mounted LiDAR to generate raw LiDAR point cloud data;

[0013] S12: Perform morphological filtering and density-based noise point removal on the original lidar point cloud data to generate denoised lidar point cloud data;

[0014] S13: performing digital elevation model extraction processing based on the stereo image pair of satellite remote sensing images to generate digital elevation model data from multi-source terrain data;

[0015] S14: Perform iterative closest point registration processing on the denoised radar point cloud data and the digital elevation model data to generate a high-precision registered point cloud;

[0016] S15: Extract the slope, aspect, curvature and elevation variation coefficient of the high-precision registration point cloud based on the digital elevation model data, and construct the terrain parameter matrix.

[0017] In one embodiment, S2 of a dynamic monitoring method for forest cultivation based on remote sensing technology provided by the present invention specifically includes the following steps:

[0018] S21: Process the terrain parameter matrix to construct a terrain characteristic vector including slope, aspect, curvature and elevation variation coefficient;

[0019] S22: Perform multi-scale classification processing on terrain feature vectors based on graph-structured neural networks to generate terrain classification results for three types of terrain units: steep slopes, gentle slopes, and gullies.

[0020] S23: Based on the terrain classification results, the echo intensity weighted algorithm is used to lower the ground point classification threshold for the lidar point cloud data in the steep slope area. The multi-echo intensity and incident angle parameters are fused to suppress point cloud missing in the lidar point cloud data in the valley area to generate canopy point cloud segmentation results for different terrain units.

[0021] In one embodiment, S3 of a remote sensing-based forest cultivation dynamic monitoring method provided by the present invention specifically includes the following steps:

[0022] S31: Perform data statistics on the canopy point cloud segmentation results, calculate the ratio of the effective canopy points to the total points in different terrain units, and obtain the penetration rate of different terrain units;

[0023] S32: Based on multiple regression analysis, the slope, aspect and elevation variation coefficients in the terrain parameter matrix and the permeability of different terrain units are modeled to generate the permeability compensation coefficient;

[0024] S33: constructing a canopy penetration compensation model based on the penetration compensation coefficient. The canopy penetration compensation model is used to dynamically correct the penetration deviation of the lidar point cloud.

[0025] S34: Perform 3D structural parameter inversion processing on the LiDAR point cloud data based on the canopy penetration compensation model to generate corrected forest 3D structural parameters.

[0026] In one embodiment, S34 of a dynamic monitoring method for forest cultivation based on remote sensing technology provided by the present invention specifically includes the following steps:

[0027] S341: Load and process the canopy penetration compensation model and canopy point cloud segmentation results based on distributed storage technology to generate a dataset to be corrected;

[0028] S342: performing spatial interpolation processing on the penetration deviation in the to-be-corrected data set based on a point cloud interpolation algorithm to generate canopy point cloud data after deviation correction;

[0029] S343: Perform three-dimensional parameter inversion processing on the bias-corrected canopy point cloud data based on the radiation transfer algorithm to generate corrected forest three-dimensional structural parameters.

[0030] In one embodiment, S4 of a dynamic monitoring method for forest cultivation based on remote sensing technology provided by the present invention specifically includes the following steps:

[0031] S41: Time series alignment of forest 3D structural parameters based on multi-temporal satellite remote sensing data to generate a temporally and spatially consistent vegetation index time series dataset;

[0032] S42: Based on time series analysis technology, trend decomposition processing is performed on the vegetation index time series dataset to extract forest growth trend characteristics and abnormal fluctuation characteristics;

[0033] S43: Based on the machine learning classification algorithm, the forest growth trend characteristics and abnormal fluctuation characteristics are integrated and analyzed to generate cultivation monitoring results. The cultivation monitoring results are used to indicate the risks of pests and diseases, fire hazards and the effects of artificial forest cultivation.

[0034] In one embodiment, S42 of a forest cultivation dynamic monitoring method based on remote sensing technology provided by the present invention specifically includes the following steps:

[0035] S421: Perform multi-scale decomposition on the vegetation index time series dataset based on wavelet transform technology to separate the low-frequency trend component and the high-frequency noise component, generating a trend component dataset and a noise component dataset;

[0036] S422: Smoothing the trend component dataset based on the Kalman filter algorithm to remove short-term environmental interference noise and generate forest growth trend characteristics;

[0037] S423: Perform anomaly detection processing on the noise component dataset based on the isolation forest algorithm, identify local mutation areas in the high-frequency noise, and generate abnormal fluctuation characteristics.

[0038] In a second aspect, the present invention provides a dynamic monitoring system for forest cultivation based on remote sensing technology, which is configured with the following modules:

[0039] The data acquisition and processing module is used to collect and preprocess lidar point cloud data and multi-source terrain data in complex terrain areas based on airborne LiDAR technology and satellite radar technology, and generate preprocessed high-precision registration point clouds and terrain parameter matrices.

[0040] The data segmentation optimization module is used to perform terrain-adaptive segmentation of high-precision registered point clouds based on the terrain parameter matrix. It combines the dynamic threshold segmentation algorithm to optimize the canopy point cloud penetration rate and obtain canopy point cloud segmentation results for different terrain units. The canopy point cloud segmentation results are used to indicate canopy cover density and individual tree parameters.

[0041] The forest parameter correction module is used to construct a canopy penetration compensation model based on the canopy point cloud segmentation results and the terrain parameter matrix. The module then corrects the lidar point cloud data based on the canopy penetration compensation model to generate corrected forest 3D structural parameters, including tree height, canopy width, and biomass.

[0042] The forest growth inversion module is used to process the corrected three-dimensional forest structure parameters based on multi-temporal remote sensing data, invert forest dynamic growth indicators through normalized vegetation index time series and machine learning models, and generate cultivation monitoring results. The cultivation monitoring results are used to indicate pest and disease risks, fire hazards and artificial forest cultivation effects.

[0043] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements any one of the above-mentioned dynamic monitoring methods for forest cultivation based on remote sensing technology.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for dynamic monitoring of forest cultivation based on remote sensing technology.

[0045] In summary, the present invention provides a dynamic monitoring method for forest cultivation based on remote sensing technology, which can significantly improve the inversion accuracy and dynamic tracking ability of forest parameters in complex terrain environments through the integration of full-chain technologies and dynamic optimization mechanisms. In response to the core problems of traditional methods in areas with undulating terrain such as steep slopes and valleys, such as uneven point cloud penetration, large canopy segmentation errors, and insufficient monitoring timeliness, the present invention uses terrain features to drive the data processing process, and through the collaborative acquisition and high-precision registration of multi-source remote sensing data, effectively eliminates the spatial interference of complex terrain on the lidar point cloud; adopts a dynamic threshold segmentation algorithm combined with a terrain parameter adaptive adjustment strategy to significantly improve the problem of misjudgment of single tree parameters caused by uneven density distribution of canopy point clouds; constructs a coupling model of terrain influencing factors and penetration compensation, systematically corrects the penetration deviation of the lidar point cloud in the vertical dimension, and improves the reliability of three-dimensional structural parameters such as tree height, crown width and biomass; based on the fusion analysis of multi-phase vegetation index and machine learning, it achieves the precise decoupling of forest growth trends and abnormal fluctuations, and strengthens the early warning capability of emergencies such as pests and diseases and fires. Compared with existing technologies, this method has formed a closed-loop monitoring system of "terrain perception-dynamic correction-multidimensional inversion". It not only breaks through the spatial limitations of complex terrain on remote sensing monitoring, but also significantly improves data utilization efficiency and result interpretability through automated processing procedures and multi-dimensional feature integration, providing highly reliable decision-making support for scenarios such as mountain afforestation and ecological restoration, while meeting the refined needs of forest resource surveys and real-time response requirements for disaster prevention and control.

[0046] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a flow chart of a method for dynamic monitoring of forest cultivation based on remote sensing technology provided in an embodiment of the present application;

[0048] Figure 2 A schematic diagram of a process for generating corrected forest three-dimensional structural parameters according to an embodiment of the present application;

[0049] Figure 3 A schematic diagram of a process for generating cultivation monitoring results provided in an embodiment of the present application;

[0050] Figure 4 This is a structural diagram of a forest cultivation dynamic monitoring system based on remote sensing technology provided in another embodiment of the present application. DETAILED DESCRIPTION

[0051] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate preferred embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the disclosure.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0053] In one embodiment, Figure 1 As shown, a method for dynamic monitoring of forest cultivation based on remote sensing technology is provided. This embodiment uses the method applied to a terminal as an example. It is understood that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0054] S1: Based on airborne LiDAR technology and satellite radar technology, the lidar point cloud data and multi-source terrain data in complex terrain areas are collected and preprocessed to generate preprocessed high-precision registration point clouds and terrain parameter matrices.

[0055] Specifically, the system uses airborne LiDAR (Light Detection and Ranging) technology and satellite radar technology to conduct comprehensive scans of complex terrain areas, acquiring precise LiDAR point cloud data and multi-source terrain data. The airborne LiDAR system, with its high-precision three-dimensional spatial information acquisition capabilities, adjusts flight parameters in real time during flight based on the undulating characteristics of the complex terrain, ensuring that laser pulses penetrate deep into vegetation-covered areas, penetrating the tree canopy, reaching the understory and ground, thereby acquiring complete forest structure information. Meanwhile, satellite radar technology, with its wide-area coverage and all-weather monitoring capabilities, provides the system with macroscopic terrain and vegetation cover data.

[0056] The preprocessing process encompasses multiple steps, including data denoising, filtering, and coordinate conversion. Denoising utilizes advanced filtering algorithms to effectively remove noise points caused by equipment errors, environmental interference, and other factors, while preserving the true terrain and vegetation reflection signals. Filtering distinguishes surface points from non-surface points based on specific rules, initially constructing a terrain outline. Coordinate conversion unifies data from different coordinate systems into a standard geographic coordinate system, laying the foundation for subsequent data fusion. After preprocessing, the system further applies a high-precision registration algorithm to precisely align the LiDAR point cloud data with multi-source terrain data, generating a high-precision registered point cloud and terrain parameter matrix. This matrix not only contains rich terrain information, such as elevation, slope, and aspect, but also incorporates vegetation distribution characteristics, providing a detailed and accurate data foundation for subsequent monitoring steps.

[0057] S2: Based on the terrain parameter matrix, the high-precision registered point cloud is subjected to terrain adaptive segmentation. The canopy point cloud penetration rate is optimized by combining the dynamic threshold segmentation algorithm to obtain the canopy point cloud segmentation results of different terrain units. The canopy point cloud segmentation results are used to indicate the canopy cover density and individual tree parameters.

[0058] Specifically, with the terrain parameter matrix as the key basis, the system conducts an in-depth analysis of the high-precision registered point cloud, fully considering the terrain variation characteristics of complex terrain areas, such as the steepness of mountains, the undulating patterns of hills, and the morphological characteristics of valleys, and finely divides the entire monitoring area into multiple terrain units with similar terrain characteristics. Within each terrain unit, the system uses a self-developed dynamic threshold segmentation algorithm to perform detailed segmentation of the canopy point cloud. This algorithm abandons the limitations of traditional fixed thresholds and instead dynamically adjusts the segmentation threshold in real time based on the specific characteristics of the terrain unit, including the slope, aspect, and coverage density and type of vegetation. For example, in areas with sparse vegetation and gentle slopes, the segmentation threshold is appropriately lowered to fully extract the detailed information of the canopy; while in areas with dense vegetation and steep slopes, the segmentation threshold is appropriately increased to avoid mis-segmentation and ensure the accurate presentation of the canopy structure.

[0059] Through this process, the system generates canopy point cloud segmentation results for different terrain units. These results not only intuitively indicate the distribution of canopy cover density but also provide a key basis for the subsequent extraction of individual tree parameters. Tree parameter extraction is based on the geometric and spatial distribution characteristics of the canopy point cloud. Through complex algorithms, the system accurately calculates key parameters such as individual tree height and crown width, providing fundamental data support for further forest growth analysis and health assessment.

[0060] S3: Based on the canopy point cloud segmentation results and the terrain parameter matrix, a canopy penetration compensation model is constructed. The lidar point cloud data is corrected based on the canopy penetration compensation model to generate corrected forest 3D structural parameters, including tree height, crown width, and biomass.

[0061] Specifically, by establishing a complex relationship model between canopy penetration and terrain factors, the system can accurately correct lidar point cloud data based on differences in canopy penetration across different terrain units. For example, in areas with dense vegetation cover, the system adjusts the distribution of point cloud data to reduce the impact of canopy obstruction on the inversion of forest 3D structural parameters. In areas with steep slopes, the density of point cloud data is increased to compensate for the reduced point cloud penetration caused by the undulating terrain. After correction, the system generates corrected 3D forest structural parameters, which include key indicators such as tree height, crown width, and biomass. Tree height is precisely calculated by analyzing the vertical distance between the top of the canopy and the ground in the corrected point cloud data; crown width is determined based on the horizontal projection of the canopy; and biomass estimation is calculated using a complex biomass model that combines parameters such as tree height, crown width, and tree density. The corrected 3D forest structural parameters more realistically reflect the growth status of the forest, providing accurate data support for subsequent dynamic monitoring of forest cultivation.

[0062] S4: Based on multi-temporal remote sensing data, the corrected three-dimensional forest structural parameters are processed, and the forest dynamic growth indicators are inverted through the normalized vegetation index time series and machine learning models to generate cultivation monitoring results. The cultivation monitoring results are used to indicate the risks of pests and diseases, fire hazards and the cultivation effects of artificial forests.

[0063] Specifically, the system calculates the Normalized Difference Vegetation Index (NDVI) for remote sensing images acquired at different times to obtain time series data of the forest vegetation index. The NDVI time series can intuitively show the changing trends of forest vegetation at different growth stages, such as the increase or decrease in vegetation coverage, the start and end time of the growing season, etc. The system combines advanced machine learning models, such as random forests and support vector machines, to conduct in-depth analysis of NDVI time series data, thereby inverting forest dynamic growth indicators. For example, by analyzing abnormal changes in vegetation indices, the system can accurately assess the risk of pests and diseases. When the NDVI value in a certain area suddenly drops or continues to be lower than normal, the system will issue a pest and disease warning; by monitoring the dryness of vegetation and the fuel load, the system can assess fire hazards and take preventive measures in advance; at the same time, the system can also evaluate the effectiveness of artificial forest cultivation, such as tree growth rate and survival rate, based on comparative analysis of forest growth indicators and expected targets. Finally, the system generates cultivation monitoring results, which present information such as forest pest and disease risks, fire hazards, and artificial forest cultivation effects in an intuitive manner, providing a scientific basis for forest cultivation management decisions, helping forestry workers optimize forest cultivation plans, and improving the management and protection of forest resources.

[0064] In summary, the present invention provides a dynamic monitoring method for forest cultivation based on remote sensing technology, which can significantly improve the inversion accuracy and dynamic tracking ability of forest parameters in complex terrain environments through the integration of full-chain technologies and dynamic optimization mechanisms. In response to the core problems of traditional methods in areas with undulating terrain such as steep slopes and valleys, such as uneven point cloud penetration, large canopy segmentation errors, and insufficient monitoring timeliness, the present invention uses terrain features to drive the data processing process, and through the collaborative acquisition and high-precision registration of multi-source remote sensing data, effectively eliminates the spatial interference of complex terrain on the lidar point cloud; adopts a dynamic threshold segmentation algorithm combined with a terrain parameter adaptive adjustment strategy to significantly improve the problem of misjudgment of single tree parameters caused by uneven density distribution of canopy point clouds; constructs a coupling model of terrain influencing factors and penetration compensation, systematically corrects the penetration deviation of the lidar point cloud in the vertical dimension, and improves the reliability of three-dimensional structural parameters such as tree height, crown width and biomass; based on the fusion analysis of multi-phase vegetation index and machine learning, it achieves the precise decoupling of forest growth trends and abnormal fluctuations, and strengthens the early warning capability of emergencies such as pests and diseases and fires. Compared with existing technologies, this method has formed a closed-loop monitoring system of "terrain perception-dynamic correction-multidimensional inversion". It not only breaks through the spatial limitations of complex terrain on remote sensing monitoring, but also significantly improves data utilization efficiency and result interpretability through automated processing procedures and multi-dimensional feature integration, providing highly reliable decision-making support for scenarios such as mountain afforestation and ecological restoration, while meeting the refined needs of forest resource surveys and real-time response requirements for disaster prevention and control.

[0065] In one embodiment, the present invention provides a method for dynamic monitoring of forest cultivation based on remote sensing technology, step S1 specifically comprising the following steps:

[0066] S11: Perform multiple scans of complex terrain areas using drone-mounted lidar to generate raw lidar point cloud data.

[0067] Specifically, the system utilizes drone-mounted LiDAR technology, enabling efficient and accurate multi-flight scanning of complex terrain. During scanning missions, the drone flies over the target forest area multiple times at a pre-set altitude, speed, and flight path, ensuring comprehensive and detailed terrain and vegetation information. The LiDAR device continuously emits high-frequency laser pulses during flight. These pulses penetrate the forest canopy, reflecting off tree branches, trunks, and the ground before returning to the sensor. This provides a wealth of three-dimensional spatial information, including tree heights, canopy dimensions, terrain undulations, and vegetation density. Data from each flight is initially timestamped and annotated with location information to ensure accurate mapping to specific locations during subsequent data integration. This pre-processed data set forms a richly detailed raw LiDAR point cloud, providing a solid foundation for subsequent in-depth analysis and processing. This ensures the system can capture highly accurate terrain and vegetation information even in complex terrain, providing a reliable data source for dynamic forest cultivation monitoring.

[0068] S12: Perform morphological filtering and density-based noise point removal on the original lidar point cloud data to generate denoised lidar point cloud data.

[0069] Specifically, the system processes the raw data using advanced morphological filtering techniques. Based on mathematical morphology theory, morphological filtering utilizes specific structuring elements to perform operations such as dilation and erosion on the point cloud data, effectively removing noise points caused by environmental interference, equipment errors, or non-target reflections. This process smoothes the point cloud data surface, reducing glitches and abrupt points, while preserving key morphological features of target objects, such as tree outlines and terrain edges. The system then applies a density-based noise removal algorithm to deeply cleanse the filtered data. This algorithm analyzes the spatial density distribution of the point cloud data to identify outliers with significant density differences from surrounding data. These outliers are often caused by random interference or non-vegetation objects such as birds and insects. Based on a pre-set density threshold, the system accurately removes these noise points while avoiding the inadvertent deletion of real vegetation or terrain data. After these two rigorous noise removal steps, the resulting denoised radar point cloud data significantly improves in quality and accuracy, laying a solid foundation for subsequent terrain analysis and vegetation parameter extraction, ensuring the system can perform further processing and analysis based on high-quality data.

[0070] S13: Perform digital elevation model extraction processing based on the stereo image pair of satellite remote sensing images to generate digital elevation model data in the multi-source terrain data.

[0071] Specifically, by acquiring stereo pairs of satellite remote sensing images, the system can accurately extract and process digital elevation models (DEMs). Satellite remote sensing images, with their large-area coverage, multi-temporal, and multi-spectral characteristics, provide a wealth of surface information. Stereo pairs are obtained from two or more images taken from different angles. Using the principle of parallax, the system can calculate the elevation information of surface points. During processing, the system first performs radiometric and geometric corrections on the remote sensing images to eliminate errors caused by factors such as atmospheric conditions, sensor characteristics, and satellite orbital position, ensuring image accuracy and consistency. Next, a feature matching algorithm identifies and matches the same-name points in the stereo pairs. These points are used to calculate the three-dimensional coordinate information of the surface, generating high-precision digital elevation model data. This data not only contains surface elevation information but also clearly reflects the microscopic relief features of the terrain, such as hills, gullies, and other topographic details.

[0072] Digital elevation model data from multi-source terrain data complements lidar point cloud data. Lidar data offers higher vertical accuracy in vegetation-covered areas, while satellite remote sensing imagery excels in covering large areas and extracting macro-terrain features. By fusing these two types of data, the system obtains more comprehensive and accurate terrain information, providing a richer terrain reference for subsequent dynamic monitoring of forest cultivation and ensuring accurate terrain analysis and vegetation monitoring even in complex terrain conditions.

[0073] S14: Perform iterative closest point registration processing on the denoised radar point cloud data and the digital elevation model data to generate a high-precision registered point cloud.

[0074] Specifically, the registration process can use the Iterative Closest Point (ICP) algorithm, which achieves accurate registration by minimizing the distance between the two data sets. Before starting the registration, the system first performs an initial registration on the denoised radar point cloud data and the digital elevation model data, using their spatial position information and attitude parameters to roughly align them to the same coordinate system. Then, the Iterative Closest Point algorithm comes into play, searching for corresponding point pairs between the two sets of data and calculating the Euclidean distance between them. Through continuous iterative optimization, the translation and rotation parameters of the point cloud data are adjusted, gradually reducing the distance between corresponding point pairs until the preset convergence accuracy threshold is reached.

[0075] Throughout the registration process, the system can also use a variety of strategies to improve the accuracy and efficiency of registration, such as using a stratified sampling method to prioritize matching significant feature points on the ground, such as the bottom of tree trunks, terrain edge points, etc. These feature points have high identifiability and stability and can effectively guide the registration process. At the same time, the system also introduces an error assessment mechanism to monitor the error changes in the registration process in real time, adjust the algorithm parameters in a timely manner, and ensure the reliability of the registration results. After this sophisticated registration process, the generated high-precision registration point cloud is not only highly consistent with the digital elevation model in terms of geometric position, but also retains rich vegetation structure information, providing an accurate and comprehensive data basis for subsequent terrain parameter extraction and forest three-dimensional structure analysis, enabling the system to more accurately analyze forest terrain characteristics and vegetation distribution, and provide strong support for forest cultivation decisions.

[0076] S15: Extract the slope, aspect, curvature and elevation variation coefficient of the high-precision registration point cloud based on the digital elevation model data, and construct the terrain parameter matrix.

[0077] Specifically, based on digital elevation model data, the system extracts key terrain parameters such as slope, aspect, curvature and elevation variation coefficient from high-precision registration point clouds, and constructs a terrain parameter matrix. Slope is an important indicator that reflects the degree of inclination of the terrain surface. Its calculation method is usually based on the elevation difference between adjacent grids in the digital elevation model data, and the slope value of each grid is calculated through trigonometric functions. The aspect indicates the direction of the slope inclination. The system determines the aspect of each grid by analyzing the trend of elevation changes, providing key information for studying the impact of terrain on vegetation distribution and growth. Curvature is a parameter that describes the degree of curvature of the terrain surface. The system obtains curvature information of the terrain surface by performing secondary difference calculations on the elevation data, which helps to identify the concave and convex features of the terrain. The elevation variation coefficient is an indicator that measures the degree of discreteness of elevation data, reflecting the complexity and degree of undulation of the terrain.

[0078] The system uses a combination of algorithms and mathematical models to accurately calculate these terrain parameters and organize them into a terrain parameter matrix. This matrix not only comprehensively reflects the topographic characteristics of complex terrain areas but also integrates closely with denoised radar point cloud data, providing rich terrain background information for subsequent dynamic monitoring of forest cultivation. This facilitates in-depth analysis of the interaction between forests and terrain, providing strong support for scientific management of forest resources and cultivation decisions.

[0079] In one embodiment, S2 of a dynamic monitoring method for forest cultivation based on remote sensing technology provided by the present invention specifically includes the following steps:

[0080] S21: Process the terrain parameter matrix and construct a terrain characteristic vector containing slope, aspect, curvature and elevation variation coefficient.

[0081] Specifically, the system first extracts key terrain parameters such as slope, aspect, curvature, and elevation variation coefficient from the terrain parameter matrix. These parameters comprehensively reflect the topographic characteristics of complex terrain areas. For example, slope reflects the degree of surface inclination, aspect indicates the direction of surface inclination, curvature reveals the curved characteristics of the terrain surface, and the elevation variation coefficient measures the degree of dispersion of elevation data. Combined, these parameters clearly depict the topography and landforms of complex terrain areas. The system then combines these parameters into a terrain feature vector, forming a complete description of the terrain characteristics of the complex terrain area, providing the basic data basis for subsequent terrain classification and analysis.

[0082] S22: Based on the graph structure neural network, the terrain feature vectors are classified at multiple scales to generate terrain classification results for three types of terrain units: steep slope, gentle slope, and valley.

[0083] Specifically, graph-structured neural networks are capable of processing complex non-Euclidean data structures and are suitable for classifying terrain feature vectors. During the classification process, the system leverages the multi-scale nature of graph-structured neural networks to analyze terrain feature vectors at different scales. At small scales, they can capture local features of the terrain, such as individual trees or slope variations within a small area. At larger scales, they can grasp the distribution and trends of the overall terrain, such as the direction of a mountain range or the extent of a large valley area.

[0084] Through comprehensive multi-scale analysis, the graph-structured neural network can more accurately identify the characteristics of different terrain units, classifying them into three categories: steep slopes, gentle slopes, and valleys. For example, steep slopes typically have larger slope values and a higher coefficient of elevation variation, while gentle slopes have smaller slopes and flatter terrain. The terrain feature vectors in valleys exhibit unique characteristics such as curvature. Ultimately, the system generates accurate terrain classification results, providing precise terrain information for subsequent analysis and processing.

[0085] S23: Based on the terrain classification results, the echo intensity weighted algorithm is used to lower the ground point classification threshold for the lidar point cloud data in the steep slope area. The multi-echo intensity and incident angle parameters are fused to suppress point cloud missing in the lidar point cloud data in the valley area to generate canopy point cloud segmentation results for different terrain units.

[0086] Specifically, in steep slope areas, the system uses an echo intensity weighting algorithm to weight the classification process of ground points using echo intensity information, thereby lowering the threshold for ground point classification. Echo intensity reflects the characteristics of the interaction between the laser and the ground surface. In steep slope areas, due to the inclined terrain, the echo intensity of the lidar point cloud has specific distribution characteristics. Through this algorithm, the system can more accurately identify ground points and non-ground points, reducing misclassification. In valley areas, the system fuses multiple echo intensities with incident angle parameters to suppress point cloud loss. Valley areas are prone to point cloud data loss due to factors such as terrain obstruction. Multiple echo intensity information can provide richer surface reflection characteristics, and the incident angle parameter takes into account the impact of the laser incident angle on the echo intensity and point cloud distribution.

[0087] By fusing these two types of information, the system can, to a certain extent, compensate for missing point clouds and improve data integrity and accuracy. Ultimately, the system generates corresponding canopy point cloud segmentation results based on the characteristics of different terrain units, providing high-quality data for subsequent forest parameter inversion and analysis.

[0088] In one embodiment, S3 of a remote sensing-based forest cultivation dynamic monitoring method provided by the present invention specifically includes the following steps:

[0089] S31: Perform data statistics on the canopy point cloud segmentation results, calculate the ratio of the effective canopy points to the total number of points in different terrain units, and obtain the penetration rate of different terrain units.

[0090] Specifically, when processing canopy point cloud segmentation results, the system comprehensively considers all terrain units, including steep slopes, gentle slopes, and gullies. For each terrain unit, the system accurately counts valid canopy points—those point cloud data points that, after detailed segmentation, are clearly identified as forest canopy. The system not only counts valid canopy points but also the total number of points within each terrain unit, including canopy points, ground points, and any other non-canopy and non-ground points. By calculating the ratio of valid canopy points to total points, the system determines the penetration rate for that terrain unit. This penetration rate clearly reflects the ability of the lidar signal to penetrate the canopy within that specific terrain unit. The system generates corresponding canopy point cloud segmentation results based on the characteristics of each terrain unit, providing high-quality data for subsequent forest parameter inversion and analysis. For example, in areas with sparse vegetation and gentle slopes, the system accurately identifies more valid canopy points, while in areas with dense vegetation and steep slopes, the system adjusts its subsequent analysis strategy based on the penetration rate calculation results. This process not only ensures the accuracy of the data, but also provides a solid foundation for subsequent modeling of terrain influencing factors.

[0091] S32: Based on multiple regression analysis, the slope, aspect and elevation variation coefficients in the terrain parameter matrix and the permeability of different terrain units are modeled as terrain influence factors to generate permeability compensation coefficients.

[0092] Specifically, multiple regression analysis can reveal the complex relationships between multiple independent variables and a single dependent variable. In this example, slope, aspect, and elevation coefficient of variation serve as independent variables, while the penetration rate of different terrain units serves as the dependent variable. By collecting a large amount of sample data, the system establishes a multiple regression model, deeply analyzing the extent and direction of the impact of each terrain parameter on penetration rate. For example, areas with steeper slopes may have higher penetration rates due to the sloping terrain, making it easier for LiDAR signals to penetrate the canopy. Meanwhile, areas with higher elevation coefficients of variation may experience complex interference effects on LiDAR signal penetration. Through multiple regression analysis, the system determines the specific relationship between these terrain parameters and penetration rate and expresses it in the form of a mathematical model. The system generates a penetration compensation coefficient that reflects the extent of the terrain's impact on penetration rate, providing a key parameter for subsequent penetration rate correction. The system dynamically adjusts the parameters of the multiple regression model based on the terrain characteristics of each terrain unit to ensure model accuracy and adaptability. This process not only improves the reliability of the model but also provides a scientific basis for the subsequent construction of a canopy penetration compensation model.

[0093] S33: Based on the penetration compensation coefficient, a canopy penetration compensation model is constructed. The canopy penetration compensation model is used to dynamically correct the penetration deviation of the lidar point cloud.

[0094] Specifically, in complex terrain, the penetration of LiDAR point clouds often exhibits certain deviations due to the complexity of topographic factors, which can affect the accuracy of subsequent inversion of forest 3D structural parameters. The canopy penetration compensation model, by introducing a penetration compensation coefficient, enables real-time adjustment and correction of the penetration of LiDAR point clouds based on varying terrain conditions. For example, in steep slopes, the model appropriately increases the estimated penetration based on the penetration compensation coefficient to compensate for potential underestimation caused by the steep terrain. In valleys, the model also adjusts the penetration based on the compensation coefficient. This dynamic correction mechanism ensures that the penetration of LiDAR point cloud data is accurately reflected in different terrain units, thereby improving the reliability of subsequent inversion results. The system dynamically adjusts the model's correction parameters based on the topographic characteristics of each unit to ensure model adaptability and accuracy. This process not only improves the model's correction effectiveness but also provides a solid foundation for subsequent inversion of 3D structural parameters.

[0095] S34: Perform 3D structural parameter inversion processing on the LiDAR point cloud data based on the canopy penetration compensation model to generate corrected forest 3D structural parameters.

[0096] Specifically, the corrected forest 3D structure parameters are generated by the following steps:

[0097] S341: Based on distributed storage technology, the canopy penetration compensation model and the canopy point cloud segmentation results are loaded and processed to generate a data set to be corrected.

[0098] Specifically, distributed storage technology achieves parallel processing and fast access to data by dividing large-scale data sets into multiple data blocks and distributing and storing these data blocks on multiple storage nodes. During the loading process, the system first divides the relevant parameters of the canopy permeability compensation model and the canopy point cloud segmentation results into multiple sub-datasets, and then distributes these sub-datasets to different storage nodes. Each storage node is responsible for loading and processing its corresponding sub-dataset, significantly improving the efficiency of data loading through parallel computing. During the integration process, the system ensures data consistency and spatial correlation between each data block, and ensures data integrity and accuracy by establishing indexing and data verification mechanisms. The system preprocesses these data, including data format conversion, coordinate system operations, etc., so that data from different sources can be processed under the same framework. In this way, the system generates a dataset to be corrected with a complete structure and detailed content, providing efficient data support for subsequent deviation correction processing.

[0099] S342: Performing spatial interpolation processing on the penetration deviation in the to-be-corrected dataset based on a point cloud interpolation algorithm to generate canopy point cloud data after deviation correction.

[0100] Specifically, the system can use interpolation algorithms such as inverse distance weighted interpolation and kriging interpolation to perform deviation correction processing on each point in the data set. Specifically, the system analyzes the penetration deviation of each point in the data set to be corrected, and identifies areas with larger deviations and areas with smaller deviations. Then, based on the deviation values of the surrounding known points, the interpolation algorithm is used to calculate the deviation correction value of the unknown point. During the interpolation process, the system fully considers spatial correlation and terrain characteristics to ensure the accuracy and rationality of the correction results. For example, in areas where the terrain changes drastically, the system uses more refined interpolation parameters to capture the impact of the terrain on the penetration deviation. At the same time, the system performs a quality assessment on the interpolation results, and ensures the reliability of the interpolation results by calculating the interpolation error and performing cross-validation. After spatial interpolation processing, the system generates canopy point cloud data after deviation correction. These data more accurately reflect the actual situation in terms of penetration, providing a more reliable data basis for the subsequent three-dimensional structural parameter inversion.

[0101] S343: Perform three-dimensional parameter inversion processing on the bias-corrected canopy point cloud data based on the radiation transfer algorithm to generate corrected forest three-dimensional structural parameters.

[0102] Specifically, the radiative transfer algorithm is a physical model that simulates the propagation of light through the vegetation canopy. The system uses this algorithm to conduct in-depth analysis of the calibrated canopy point cloud data to extract the three-dimensional structural information of the forest. During the inversion process, the system first performs a beam tracing simulation on the calibrated point cloud data, analyzing the beam's propagation path, reflection, and absorption within the canopy. By calculating parameters such as the beam's path length, number of reflections, and energy attenuation, the system can infer the canopy's optical and structural characteristics. Combining the calibrated point cloud data, the system further calculates key parameters such as tree height, crown width, and biomass. Tree height is determined by analyzing the beam path length between the top of the canopy and the ground; crown width is calculated based on the horizontal beam coverage of the canopy; and biomass is estimated by combining parameters such as tree height, crown width, and tree density using the biomass estimation formula in the radiative transfer model. The system performs detailed statistical analysis on these parameters to generate the calibrated three-dimensional forest structural parameters. These parameters can not only more accurately reflect the actual growth conditions and structural characteristics of the forest, but also provide reliable data support for dynamic monitoring of forest cultivation, helping forestry workers better understand the growth of the forest and make scientific and reasonable forest management decisions.

[0103] The above-mentioned dynamic monitoring method of forest cultivation based on remote sensing technology systematically solves the core problem of insufficient accuracy of canopy parameter inversion caused by uneven distribution of lidar point cloud penetration in complex terrain environments by integrating terrain characteristics and dynamic correction mechanism of penetration rate. In response to the problems of steep slopes and valley areas caused by terrain undulations and the mis-segmentation of canopy structure in traditional methods, the present invention proposes a full-process correction strategy from data statistics to model construction: by quantifying the proportion of effective canopy point clouds in different terrain units, a dynamic correlation model of penetration rate and terrain factors such as slope and aspect is established, breaking through the limitation of traditional static threshold segmentation that is not adaptable to terrain heterogeneity; based on multivariate regression analysis, a penetration rate compensation model driven by terrain influencing factors is constructed to achieve spatial adaptive correction of point cloud penetration rate deviations, effectively eliminating the interference of terrain undulations on the analysis of canopy vertical structure; combining the principle of radiation transmission with the three-dimensional parameter inversion of the corrected point cloud data, the reliability and spatial consistency of key parameters such as tree height, crown width and biomass are significantly improved. Compared with existing technologies, this method forms a closed-loop optimization mechanism of "penetration quantification-dynamic modeling-bias correction-parameter inversion" through the deep integration of terrain characteristics and penetration compensation. It can not only adapt to the spatial heterogeneity characteristics of complex terrain, but also improve the refinement level of forest three-dimensional structure analysis through multi-dimensional data coupling, providing highly reliable data support for mountain afforestation project acceptance, forest carbon sink accounting and ecological restoration effect evaluation. At the same time, it lays a foundation for precise analysis of dynamic early warning scenarios such as pest and disease monitoring and fire hazard identification, and comprehensively enhances the technical universality and engineering applicability of forest resource monitoring in complex terrain areas.

[0104] In one embodiment, S4 of a dynamic monitoring method for forest cultivation based on remote sensing technology provided by the present invention specifically includes the following steps:

[0105] S41: Perform time series alignment processing on forest 3D structural parameters based on multi-temporal satellite remote sensing data to generate a temporally and spatially consistent vegetation index time series dataset.

[0106] Specifically, multi-temporal satellite remote sensing data covers vegetation growth information at different points in time. The system uses a time series alignment algorithm to align these data across time, ensuring that data from each point in time can be compared and analyzed within the same spatiotemporal framework. During the alignment process, the system uses interpolation algorithms to fill in information gaps within the data collection interval, taking into account factors such as the temporal resolution and orbital repetition period of the satellite data. This generates a temporally and spatially consistent vegetation index time series dataset. These datasets not only capture changes in vegetation indices but also integrate information from multiple sources, such as topography and meteorology, providing comprehensive data support for subsequent forest growth trend analysis. For example, when processing the data, the system analyzes changes in vegetation indices across seasons and years, incorporating factors such as terrain slope and aspect to assess long-term trends and seasonal patterns in vegetation growth. In this way, the system accurately captures the dynamic changes in forest vegetation over time, providing a solid data foundation for subsequent monitoring and analysis.

[0107] S42: Based on time series analysis technology, trend decomposition processing is performed on the vegetation index time series dataset to extract forest growth trend characteristics and abnormal fluctuation characteristics.

[0108] Specifically, through trend decomposition processing, the system can clearly separate the long-term trend, seasonal changes and random fluctuation components of the forest vegetation index. The long-term trend reflects the overall growth status of the forest, the seasonal changes reflect the regular growth pattern of vegetation with the change of seasons, and the random fluctuation component may include the impact of abnormal events such as pests and diseases and fires. The system analyzes these characteristics in detail, identifies the time of occurrence, duration and impact of abnormal fluctuations, and compares them with historical data to evaluate the stability and health of forest growth. Preferably, the forest growth trend characteristics and abnormal fluctuation characteristics are obtained through the following steps:

[0109] S421: Perform multi-scale decomposition on the vegetation index time series dataset based on wavelet transform technology to separate the low-frequency trend component and the high-frequency noise component, and generate a trend component dataset and a noise component dataset.

[0110] Specifically, the wavelet transform decomposes signals into components of varying scales and frequencies. By appropriately selecting wavelet basis functions and decomposition scales, the system decomposes vegetation index time series datasets into low-frequency trend components and high-frequency noise components. The low-frequency component primarily reflects the long-term trend of the vegetation index, while the high-frequency component contains short-term fluctuations and noise information. During the decomposition process, the system employs multiscale analysis methods to extract features from different scales, ensuring the accuracy and reliability of the decomposition results. Ultimately, the system generates trend and noise component datasets, providing a foundation for further analysis and processing.

[0111] S422: Smoothing the trend component dataset based on the Kalman filter algorithm to remove short-term environmental interference noise and generate forest growth trend characteristics.

[0112] Specifically, the Kalman filter algorithm is a recursive estimation algorithm that can effectively remove short-term environmental interference noise from the data. By establishing a state-space model, the system treats the vegetation index in the trend component dataset as a state variable and uses the prediction and update steps of the Kalman filter to dynamically estimate the true value of the vegetation index. In the prediction step, the system predicts the vegetation index value at the next moment based on the state transition model; in the update step, the predicted value is corrected based on the actual observed value to obtain a more accurate estimate. In this way, the Kalman filter algorithm can effectively track the changing trend of the vegetation index and remove short-term environmental interference noise, ultimately generating a smooth forest growth trend feature. The system evaluates and verifies the processed results to ensure that the generated forest growth trend feature can truly reflect the long-term growth status of the forest.

[0113] S423: Perform anomaly detection processing on the noise component dataset based on the isolation forest algorithm, identify local mutation areas in the high-frequency noise, and generate abnormal fluctuation characteristics.

[0114] Specifically, the isolation forest algorithm is an unsupervised anomaly detection method based on random forests that can effectively identify local mutation areas in data. The system analyzes each data point in the noise component dataset by constructing an isolation forest model. The isolation forest algorithm constructs multiple isolated trees by randomly selecting features and split points. The depth of each data point in the isolated tree reflects its degree of abnormality. Based on the anomaly scores of the data points, the system identifies those local mutation areas with significant differences. These local mutation areas may correspond to abnormal events in the forest, such as outbreaks of pests and diseases or the impact of fire. The system conducts detailed analysis and annotation of the identified abnormal fluctuation features, and generates an abnormal fluctuation feature dataset for subsequent analysis. In this way, the system can promptly detect abnormal conditions in the forest growth process and provide important early warning information for forest cultivation monitoring.

[0115] S43: Based on the machine learning classification algorithm, the forest growth trend characteristics and abnormal fluctuation characteristics are integrated and analyzed to generate cultivation monitoring results. The cultivation monitoring results are used to indicate the risks of pests and diseases, fire hazards and the effects of artificial forest cultivation.

[0116] Specifically, the system uses growth trend characteristics and abnormal fluctuation characteristics as input variables and analyzes them using a trained classification model. For example, the random forest algorithm constructs multiple decision trees, comprehensively considering the importance and weighting of multiple features to classify forest health. For pest and disease risk, the system identifies features such as abnormal declines in vegetation indices and deviations from normal growth trends. Combining geographic information and meteorological data, it predicts the potential locations and severity of pests and diseases. For fire hazards, the system analyzes indicators such as vegetation dryness and fuel load, integrating multi-source data to assess the risk level of fire. For plantation cultivation effectiveness, a comprehensive evaluation is conducted based on indicators such as the long-term growth trend and stability of the vegetation index, combined with cultivation objectives. Ultimately, the system generates intuitive cultivation monitoring results, displaying forest health and growth trends through a visual interface. This provides forestry managers with scientific decision-making support, helping them take timely measures to optimize forest cultivation plans and improve the management and protection of forest resources. For example, the system can generate pest and disease risk maps, highlighting high-risk areas so that foresters can take preventive measures in advance. In areas with high fire potential, fire prevention resources can be deployed in advance to reduce the likelihood of fire. In this way, the system can not only help improve the management efficiency of forest resources, but also provide strong guarantees for the sustainable development of forests.

[0117] In summary, the forest cultivation dynamic monitoring method based on remote sensing technology provided by the present invention systematically solves the shortcomings of traditional forest monitoring methods in dynamic tracking, abnormal warning and cultivation effect evaluation through multi-temporal data fusion and intelligent analysis technology. In response to the problems in the existing technology such as mismatched spatiotemporal resolution of multi-source remote sensing data, difficulty in decoupling growth trends from sudden disturbances, and low efficiency of manual interpretation, the present invention constructs a full-process dynamic monitoring mechanism of "data alignment-trend analysis-intelligent decision-making": through time series alignment processing of multi-phase satellite remote sensing data, the spatiotemporal benchmark deviations caused by factors such as seasonal changes and sensor differences are eliminated, and a highly consistent vegetation index data set is generated, providing a reliable data basis for long-term trend analysis; based on time series decomposition technology, the long-term regular trends of forest growth and short-term abnormal fluctuation signals are separated, and the influence of natural growth patterns and sudden interference events such as pests and diseases, fires, etc. is accurately distinguished, breaking through the limitations of traditional methods with high misjudgment rate and poor timeliness under noise interference; combined with machine learning algorithms, multi-dimensional features are fused and modeled, and the coupling relationship between vegetation index and three-dimensional structural parameters in complex terrain areas is adaptively learned to achieve integrated analysis of early hidden feature extraction of pests and diseases, location of fire hazard hotspots, and quantitative evaluation of artificial forest growth effects.

[0118] Compared with static monitoring and manual interpretation methods, this method significantly improves the perception sensitivity and result interpretability of forest dynamic changes through the deep combination of collaborative correction of time series data and machine intelligent decision-making. It can not only capture tiny abnormal fluctuations in the physiological state of the canopy, but also provide continuous and objective data support for the effectiveness evaluation of different forest management measures, thereby making up for the defects of traditional monitoring such as long cycle, high cost and lack of refinement, and providing real-time and intelligent technical means for forest resource protection, ecological restoration project management and disaster emergency response, and comprehensively promoting the digital transformation of forestry management from "post-event disposal" to "pre-event warning-in-event regulation".

[0119] Preferably, if Figure 4 As shown, the present invention provides a forest cultivation dynamic monitoring system 500 based on remote sensing technology, which is configured with the following modules:

[0120] The data acquisition and processing module 510 is used to collect and preprocess the lidar point cloud data and multi-source terrain data in complex terrain areas based on airborne LiDAR technology and satellite radar technology, and generate preprocessed high-precision registration point clouds and terrain parameter matrices.

[0121] The data segmentation optimization module 520 is used to perform terrain-adaptive segmentation on the high-precision registered point cloud based on the terrain parameter matrix, and optimize the canopy point cloud penetration rate by combining a dynamic threshold segmentation algorithm to obtain canopy point cloud segmentation results for different terrain units. The canopy point cloud segmentation results are used to indicate canopy cover density and individual tree parameters;

[0122] The forest parameter correction module 530 is used to construct a canopy penetration compensation model based on the canopy point cloud segmentation results and the terrain parameter matrix, and to correct the lidar point cloud data based on the canopy penetration compensation model to generate corrected forest 3D structural parameters, which include tree height, canopy width, and biomass.

[0123] The forest growth inversion module 540 is used to process the corrected three-dimensional forest structural parameters based on multi-temporal remote sensing data, invert forest dynamic growth indicators through the normalized vegetation index time series and machine learning model, and generate cultivation monitoring results. The cultivation monitoring results are used to indicate the risk of pests and diseases, fire hazards and the effect of artificial forest cultivation.

[0124] In summary, the forest cultivation dynamic monitoring system based on remote sensing technology provided by the present invention can significantly improve the accuracy of forest parameter inversion and dynamic tracking capabilities in complex terrain environments through full-chain technology integration and dynamic optimization mechanism. In response to the core problems of traditional methods such as uneven point cloud penetration, large canopy segmentation errors, and insufficient monitoring timeliness in undulating terrain areas such as steep slopes and valleys, the present invention uses terrain characteristics to drive the data processing process, and through the collaborative acquisition and high-precision registration of multi-source remote sensing data, effectively eliminates the spatial interference of complex terrain on the lidar point cloud; adopts a dynamic threshold segmentation algorithm combined with a terrain parameter adaptive adjustment strategy to significantly improve the problem of single tree parameter misjudgment caused by uneven canopy point cloud density distribution; constructs a terrain influencing factor and penetration compensation coupling model to systematically correct the penetration deviation of the lidar point cloud in the vertical dimension, and improves the reliability of three-dimensional structural parameters such as tree height, crown width and biomass; based on the fusion analysis of multi-phase vegetation index and machine learning, it can achieve accurate decoupling of forest growth trends and abnormal fluctuations, and enhance the early warning capability of emergencies such as pests and diseases and fires. Compared with the existing technology, the dynamic monitoring system provided by the present invention forms a closed-loop monitoring system of "terrain perception-dynamic correction-multidimensional inversion", which not only breaks through the spatial limitations of complex terrain on remote sensing monitoring, but also significantly improves data utilization efficiency and result interpretability through automated processing procedures and multi-dimensional feature integration, providing highly reliable decision-making support for scenarios such as mountain afforestation and ecological restoration, while meeting the refined needs of forest resource surveys and real-time response requirements for disaster prevention and control.

[0125] Preferably, the data acquisition and processing module 510 is configured with the following units:

[0126] A raw data generation unit, used to perform multiple scans of complex terrain areas based on the UAV-mounted LiDAR to generate raw LiDAR point cloud data;

[0127] A denoising data generation unit is used to perform morphological filtering and density-based noise point removal on the original lidar point cloud data to generate denoised lidar point cloud data;

[0128] An elevation model data generation unit is used to extract and process digital elevation models based on stereo image pairs of satellite remote sensing images, and generate digital elevation model data from multi-source terrain data;

[0129] A registration point cloud generation unit is used to perform iterative closest point registration processing on the denoised radar point cloud data and the digital elevation model data to generate a high-precision registration point cloud;

[0130] The terrain parameter matrix construction unit is used to extract the slope, aspect, curvature and elevation variation coefficient of the high-precision registration point cloud based on the digital elevation model data, and construct the terrain parameter matrix.

[0131] Preferably, the data segmentation optimization module 520 is configured with the following units:

[0132] A terrain characteristic vector construction unit is used to process the terrain parameter matrix and construct a terrain characteristic vector including slope, aspect, curvature and elevation variation coefficient;

[0133] A terrain classification result generation unit is used to perform multi-scale classification processing on terrain feature vectors based on a graph structure neural network, and generate terrain classification results for three types of terrain units: steep slope, gentle slope, and gully;

[0134] The canopy point cloud segmentation result generation unit uses the echo intensity weighted algorithm to lower the ground point classification threshold for the lidar point cloud data in the steep slope area according to the terrain classification results, and fuses the multi-echo intensity and incident angle parameters of the lidar point cloud data in the valley area to suppress point cloud missing, thereby generating canopy point cloud segmentation results for different terrain units.

[0135] Preferably, the forest parameter correction module 530 is configured with the following units:

[0136] The penetration rate calculation unit is used to perform data statistics on the canopy point cloud segmentation results, calculate the ratio of the number of effective canopy points to the total number of points in different terrain units, and obtain the penetration rate of different terrain units;

[0137] A compensation coefficient generating unit is used to model the terrain influence factors based on the slope, aspect and elevation variation coefficient in the terrain parameter matrix and the penetration rate of different terrain units based on multiple regression analysis, and generate a penetration rate compensation coefficient;

[0138] A compensation model construction unit is used to construct a canopy penetration compensation model based on the penetration compensation coefficient, and the model is used to dynamically correct the penetration deviation of the lidar point cloud;

[0139] The structural parameter generation unit is used to perform three-dimensional structural parameter inversion processing on the lidar point cloud data based on the canopy penetration compensation model to generate corrected forest three-dimensional structural parameters.

[0140] Preferably, the structural parameter generation unit includes a to-be-corrected generation subunit, a canopy correction subunit, and a forest parameter generation subunit. The to-be-corrected generation subunit is used to load and process the canopy permeability compensation model and canopy point cloud segmentation results based on distributed storage technology to generate a to-be-corrected dataset; the canopy correction subunit is used to perform spatial interpolation processing on the permeability deviation in the to-be-corrected dataset based on a point cloud interpolation algorithm to generate canopy point cloud data after deviation correction; and the forest parameter generation subunit is used to perform three-dimensional parameter inversion processing on the deviation-corrected canopy point cloud data based on a radiation transfer algorithm to generate corrected three-dimensional forest structural parameters.

[0141] Preferably, the forest growth inversion module 540 is configured with the following units:

[0142] A dataset generation unit is used to perform time series alignment processing on forest three-dimensional structural parameters based on multi-temporal satellite remote sensing data to generate a temporally and spatially consistent vegetation index time series dataset;

[0143] A feature extraction unit is used to perform trend decomposition processing on the vegetation index time series data set based on time series analysis technology to extract forest growth trend characteristics and abnormal fluctuation characteristics;

[0144] The monitoring result generation unit is used to fuse and analyze the forest growth trend characteristics and abnormal fluctuation characteristics based on the machine learning classification algorithm to generate cultivation monitoring results. The cultivation monitoring results are used to indicate the risks of pests and diseases, fire hazards and the effects of artificial forest cultivation.

[0145] Preferably, the monitoring result generation unit includes a component generation subunit, a growth trend extraction subunit, and an abnormal fluctuation generation subunit. The component generation subunit is used to perform multi-scale decomposition processing on the vegetation index time series data set based on wavelet transform technology, separate low-frequency trend components from high-frequency noise components, and generate trend component data sets and noise component data sets; the growth trend extraction subunit is used to perform smoothing processing on the trend component data set based on the Kalman filter algorithm, remove short-term environmental interference noise, and generate forest growth trend characteristics; the abnormal fluctuation generation subunit is used to perform anomaly detection processing on the noise component data set based on the isolation forest algorithm, identify local mutation areas in high-frequency noise, and generate abnormal fluctuation characteristics.

[0146] In one embodiment, the present application also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned dynamic monitoring method of forest cultivation based on remote sensing technology when executing the computer program.

[0147] In one embodiment, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned dynamic monitoring method for forest cultivation based on remote sensing technology.

[0148] In the description of this specification, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0149] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0150] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily conceive of various modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should be included within the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A dynamic monitoring method for forest cultivation based on remote sensing technology, characterized in that: The following steps are involved: S1: Based on airborne LiDAR technology and satellite radar technology, the lidar point cloud data and multi-source terrain data in complex terrain areas are collected and preprocessed to generate high-precision registered point clouds and terrain parameter matrices after preprocessing; S2: performing terrain adaptive segmentation on the high-precision registered point cloud based on the terrain parameter matrix, optimizing the canopy point cloud penetration rate in combination with a dynamic threshold segmentation algorithm, and obtaining canopy point cloud segmentation results for different terrain units. The canopy point cloud segmentation results are used to indicate canopy cover density and single tree parameters; S3: constructing a canopy penetration compensation model based on the canopy point cloud segmentation result and the terrain parameter matrix, and performing correction processing on the lidar point cloud data based on the canopy penetration compensation model to generate corrected forest three-dimensional structural parameters, wherein the forest three-dimensional structural parameters include tree height, canopy width, and biomass; S4: The corrected three-dimensional forest structural parameters are processed based on multi-temporal remote sensing data, and the forest dynamic growth indicators are inverted through the normalized vegetation index time series and the machine learning model to generate cultivation monitoring results. The cultivation monitoring results are used to indicate the risks of pests and diseases, fire hazards and the effects of artificial forest cultivation.

2. The method according to claim 1, characterized in that Said S1 comprises: S11: Perform multiple scans of complex terrain areas using UAV-mounted LiDAR to generate raw LiDAR point cloud data; S12: performing morphological filtering and density-based noise point removal processing on the original lidar point cloud data to generate denoised lidar point cloud data; S13: performing digital elevation model extraction processing based on the stereo image pair of satellite remote sensing images to generate digital elevation model data from multi-source terrain data; S14: performing iterative closest point registration processing on the denoised radar point cloud data and the digital elevation model data to generate a high-precision registered point cloud; S15: Extracting the slope, aspect, curvature and elevation variation coefficient of the high-precision registration point cloud based on the digital elevation model data, and constructing a terrain parameter matrix.

3. The method according to claim 1, characterized in that The S2 includes: S21: Processing the terrain parameter matrix to construct a terrain feature vector including slope, aspect, curvature and elevation variation coefficient; S22: performing multi-scale classification processing on the terrain feature vector based on a graph structure neural network to generate terrain classification results of three types of terrain units: steep slope, gentle slope, and gully; S23: Based on the terrain classification results, the echo intensity weighted algorithm is used to lower the ground point classification threshold for the lidar point cloud data in the steep slope area, and the multi-echo intensity and incident angle parameters are fused to suppress point cloud missing for the lidar point cloud data in the valley area to generate canopy point cloud segmentation results for different terrain units.

4. The method according to claim 1, wherein The S3 includes: S31: performing data statistics on the canopy point cloud segmentation results, calculating the ratio of the number of effective canopy points to the total number of points in different terrain units, and obtaining the penetration rate of different terrain units; S32: performing terrain influence factor modeling on the slope, aspect, and elevation variation coefficients in the terrain parameter matrix and the penetration rates of different terrain units based on multiple regression analysis to generate a penetration rate compensation coefficient; S33: constructing a canopy penetration compensation model according to the penetration compensation coefficient, wherein the canopy penetration compensation model is used to dynamically correct the penetration deviation of the lidar point cloud; S34: Performing three-dimensional structural parameter inversion processing on the lidar point cloud data based on the canopy penetration compensation model to generate corrected forest three-dimensional structural parameters.

5. The method according to claim 4, characterized in that The S34 includes: S341: Loading and processing the canopy penetration compensation model and the canopy point cloud segmentation result based on distributed storage technology to generate a data set to be corrected; S342: performing spatial interpolation processing on the penetration deviation in the to-be-corrected data set based on a point cloud interpolation algorithm to generate canopy point cloud data after deviation correction; S343: Performing three-dimensional parameter inversion processing on the bias-corrected canopy point cloud data based on a radiation transfer algorithm to generate corrected three-dimensional forest structural parameters.

6. The method according to claim 1, characterized in that The S4 includes: S41: performing time series alignment processing on the three-dimensional forest structure parameters based on multi-temporal satellite remote sensing data to generate a temporally and spatially consistent vegetation index time series dataset; S42: performing trend decomposition processing on the vegetation index time series data set based on time series analysis technology to extract forest growth trend characteristics and abnormal fluctuation characteristics; S43: Based on the machine learning classification algorithm, the forest growth trend characteristics and the abnormal fluctuation characteristics are fused and analyzed to generate a cultivation monitoring result, which is used to indicate the risk of pests and diseases, fire hazards and the effect of artificial forest cultivation.

7. The method according to claim 6, characterized in that The S42 includes: S421: performing multi-scale decomposition processing on the vegetation index time series dataset based on wavelet transform technology to separate the low-frequency trend component and the high-frequency noise component to generate a trend component dataset and a noise component dataset; S422: Smoothing the trend component dataset based on a Kalman filter algorithm to remove short-term environmental interference noise and generate forest growth trend features; S423: Perform anomaly detection processing on the noise component dataset based on the isolation forest algorithm, identify local mutation areas in the high-frequency noise, and generate abnormal fluctuation features.

8. The forest cultivation dynamic monitoring system based on remote sensing technology is characterized by: The system comprises: The data acquisition and processing module is used to collect and preprocess lidar point cloud data and multi-source terrain data in complex terrain areas based on airborne LiDAR technology and satellite radar technology, and generate preprocessed high-precision registration point clouds and terrain parameter matrices. a data segmentation optimization module for performing terrain-adaptive segmentation on the high-precision registered point cloud based on the terrain parameter matrix, optimizing the canopy point cloud penetration rate in combination with a dynamic threshold segmentation algorithm, and obtaining canopy point cloud segmentation results for different terrain units. The canopy point cloud segmentation results are used to indicate canopy cover density and individual tree parameters; a forest parameter correction module, configured to construct a canopy penetration compensation model based on the canopy point cloud segmentation results and the terrain parameter matrix, and to perform correction processing on the lidar point cloud data based on the canopy penetration compensation model to generate corrected forest three-dimensional structural parameters, wherein the forest three-dimensional structural parameters include tree height, canopy width, and biomass; The forest growth inversion module is used to process the corrected forest three-dimensional structural parameters based on multi-temporal remote sensing data, invert forest dynamic growth indicators through normalized vegetation index time series and machine learning models, and generate cultivation monitoring results. The cultivation monitoring results are used to indicate pest and disease risks, fire hazards and artificial forest cultivation effects.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images