A method for identifying damage of a bamboo forest caused by snow disaster based on a differentiated point cloud encryption strategy

Abstract

The present application relates to the technical field of forestry disaster monitoring, in particular to a method for identifying Phyllostachys edulis forest snow disaster damage based on a differentiated point cloud encryption strategy; the present application reconstructs the three-dimensional structure of the bamboo forest by fitting the bamboo stem axis and image post-processing, identifies different types of damage and generates virtual point clouds to fill in the missing areas of the scan, effectively improving the accuracy of the bamboo stem axis; through hierarchical reconstruction and principal component analysis technology, the image expression of the bamboo stem and crown is enhanced, providing more detailed damage feature maps to help accurately assess the disaster situation. In addition, the targeted image processing based on the damage type further improves the accuracy and reliability of the damage analysis. Overall, the present application provides efficient and accurate technical support for post-disaster assessment and ecological restoration of bamboo forests.

Description

Technical Field

[0001] This invention relates to the field of forestry disaster monitoring technology, specifically to a method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy. Background Technology

[0002] With the increasing frequency of climate change and extreme weather events, bamboo forests are suffering more severe damage from natural disasters such as snowstorms and windstorms. Existing point cloud processing methods have significant limitations when dealing with areas where bamboo trunks are missing, leading to incomplete disaster damage assessment results. Airborne LiDAR, as a highly efficient remote sensing technology, can quickly acquire three-dimensional point cloud data of large areas of terrain and vegetation, with high spatial resolution and accuracy, and is widely used in forest resource monitoring, disaster assessment, and other fields.

[0003] However, when lidar scans in bamboo forests, the occlusion of bamboo leaves and canopies often leads to missing point cloud data for the bamboo trunks. Existing bamboo forest damage assessment methods based on point cloud processing often neglect the filling and accurate reconstruction of missing areas on the bamboo trunks, resulting in incomplete or highly erroneous damage assessment results. Therefore, how to effectively utilize airborne lidar point cloud data to specifically identify different types of bamboo forest damage and solve the problem of missing point cloud data has become a key technical challenge in current bamboo forest damage assessment.

[0004] To address the aforementioned issues, it is necessary to propose a method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing an airborne lidar method for identifying snow disaster damage in bamboo forests based on a differentiated point cloud encryption strategy (i.e., using different point cloud generation methods according to different disaster damage types). The innovation of this invention lies in solving the problems of missing bamboo forest point cloud data and low disaster damage identification accuracy by combining hierarchical reconstruction and principal component analysis techniques with targeted image processing methods based on disaster damage types.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy includes the following steps:

[0008] Step 1: Data Acquisition and Input;

[0009] Receive point cloud data from airborne lidar;

[0010] The airborne lidar point cloud data includes the coordinates of scanning points on the bamboo surface and the laser reflection intensity, and records the average ground elevation; each scanning point on the bamboo surface has a unique laser reflection intensity.

[0011] Step 2: Data preprocessing;

[0012] The collected point cloud data is preprocessed, including data noise reduction, ground threshold separation and normalization, to obtain the normalized result of laser reflection intensity.

[0013] Gaussian filtering was applied to the coordinates of all collected bamboo surface scanning points, and then threshold separation and normalization were performed on the laser reflection intensity of each bamboo surface point.

[0014] For the coordinates of the scanned points on the bamboo surface, Gaussian filtering is used to calculate the Gaussian filter output of the coordinates of each scanned point on the bamboo surface.

[0015] In a preferred embodiment of the present invention, the Gaussian filter output of the coordinates of each bamboo surface scanning point is subjected to elevation threshold separation to obtain the above-ground part and the underground part. The laser reflection intensity of the above-ground part is normalized, and the part whose z-axis component is greater than the average ground elevation is recorded as the above-ground part, and the part whose z-axis component is less than or equal to the average ground elevation is recorded as the underground part.

[0016] For the above-ground portion, the laser reflection intensity corresponding to the original coordinate point of the Gaussian filter output of the scanning point coordinates of each bamboo surface is obtained, and then mapped to a fixed range to obtain the laser reflection intensity normalization result.

[0017] Step 3: Fitting the bamboo trunk axis;

[0018] Based on the normalized processing results of laser reflection intensity, the missing area is filled by fitting the bamboo stem axis through hierarchical reconstruction and principal component analysis.

[0019] The system is divided into layers at preset height intervals, and point cloud data matrices for each height layer are generated based on the laser reflection intensity normalization results. PCA principal component analysis is performed on the point cloud data matrix results for each height layer to calculate the eigenvectors of the point cloud covariance matrix. Discrete bamboo stem segments are generated based on the eigenvectors of the covariance matrix, and the bamboo stem axis is fitted. Virtual cloud points with preset density are generated based on the bamboo stem axis to fill the missing areas at the base of the bamboo stem.

[0020] Obtain the maximum value of the z-axis component of the ground portion from the Gaussian filtered output of all bamboo surface scanning point coordinates, and calculate the height range of each height layer. Classify the data according to the height range of each height layer, and assign the normalized laser reflection intensity results to the corresponding height layer.

[0021] In a preferred embodiment of the present invention, point cloud data matrices for each height layer are generated based on the laser reflection intensity normalization result. The first three columns of the point cloud data matrix represent the x-axis, y-axis, and z-axis components of the coordinates corresponding to the Gaussian filtered output of all bamboo surface scanning points contained in that height layer, respectively; the fourth column of the point cloud data matrix represents the laser reflection intensity normalization result corresponding to each coordinate.

[0022] In a preferred embodiment of the present invention, principal component analysis is performed on the point cloud data matrix of each height layer to calculate the covariance matrix of each height layer and solve for its corresponding eigenvalues ​​and eigenvectors; the eigenvector corresponding to the largest eigenvalue is selected as the axial direction of the bamboo trunk to obtain the bamboo trunk segments corresponding to each height layer; the bamboo trunk axis is fitted based on the bamboo trunk segments of each height layer, and the specific process is as follows:

[0023] By detecting spatially adjacent segments through connected component analysis, the endpoints of bamboo stem segments corresponding to the upper and lower height layers are obtained at the interfaces of each height layer. Segments between upper and lower layers whose x-axis or y-axis components of the starting and ending points are less than a preset threshold are marked as adjacent segments. The endpoints of bamboo stem segments in the upper and lower height layers with the smallest distance among the adjacent segments are located, and their midpoint is taken as the fitting point of the bamboo stem axis. The fitting points of all bamboo stem axes are connected to obtain the fitted bamboo stem axis.

[0024] As a preferred embodiment of the present invention, virtual cloud points with a preset density are generated based on the fitted bamboo trunk axis.

[0025] Step 4: Segmentation feature fusion and image post-processing;

[0026] Post-processing based on the fitted bamboo trunk axis enhances the image representation containing disaster damage type features.

[0027] The specific process for matching disaster loss types is as follows:

[0028] As a preferred embodiment of the present invention, if it is identified that the fitted bamboo trunk axis has no axis division, and the angle between the highest layer in the fitted bamboo trunk axis and the ground xoy plane is less than the first preset threshold angle, then it is determined to be a bent bamboo.

[0029] If the fitted bamboo stem axis is found to have no axis branching and there is a missing axis, it is determined to be a broken bamboo.

[0030] If the fitted bamboo stem axis is found to have axis branches, and the angle between the axis branches is greater than the second preset threshold angle, it is determined to be a split bamboo.

[0031] If it is detected that the angle between the fitted bamboo trunk axis and the ground xoy plane is less than the third preset threshold angle at the root position, it is determined to be a turned bamboo.

[0032] As a preferred embodiment of the present invention, the feature images of bamboo trunks and canopies at each layer are post-processed in parallel based on the disaster damage type to enhance the image representation containing disaster damage type features and reconstruct the three-dimensional structure of the entire bamboo forest. The specific process is as follows:

[0033] In a preferred embodiment of the present invention, parallel image post-processing is performed on each height layer to generate a three-dimensional structural feature map of the entire bamboo forest. For bent bamboo, the axis is extended; for broken bamboo, the fracture surface with a curvature greater than a preset threshold is located, extended along the normal direction, and random noise is added to increase the point cloud density of the fracture surface to enhance the jagged edges; for split bamboo, branching point clouds are generated at the branching points of the bamboo trunk to reconstruct the fragmented shape of the bamboo fibers; for bamboo with turned stumps, a compensation point cloud with an inclination angle multiplied by a preset density coefficient is generated at the root.

[0034] Compared with the prior art, the beneficial effects of the present invention are:

[0035] This invention provides an airborne lidar method for identifying snow disaster damage in bamboo forests based on a differentiated point cloud encryption strategy. This method can accurately identify different types of damage (such as bent bamboo, broken bamboo, split bamboo, and overturned bamboo). By fitting the bamboo trunk axis and performing image post-processing, the three-dimensional structure of the bamboo forest is reconstructed, and virtual point clouds are generated to fill in the missing areas of the scan, effectively improving the accuracy of the bamboo trunk axis. Layered reconstruction and principal component analysis techniques enhance the image representation of the bamboo trunk and canopy, providing more refined damage feature maps to help accurately assess the disaster situation. Furthermore, targeted image processing based on damage type (such as axis extension, fracture surface enhancement, and bifurcation point cloud generation) further improves the accuracy and reliability of damage analysis. Overall, this invention provides efficient and accurate technical support for post-disaster assessment and ecological restoration of bamboo forests. Attached Figure Description

[0036] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings:

[0037] Figure 1 This is a flowchart of a method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, as proposed in an embodiment of the present invention.

[0038] Figure 2 This is a schematic diagram of bamboo stem segment fitting proposed in the embodiments of the present invention;

[0039] Figure 3 This is a schematic diagram of disaster damage type matching proposed in the embodiments of the present invention;

[0040] Figure 4 This is a three-dimensional structural feature diagram of the entire bamboo forest as presented in the embodiments of the present invention;

[0041] Figure 5 This is a schematic diagram of the snow disaster damage identification process for bamboo forests proposed in the embodiments of the present invention. Detailed Implementation

[0042] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] First embodiment:

[0044] Please see Figure 1 As shown, a method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy includes the following steps:

[0045] Step 1: Data Acquisition and Input;

[0046] Receive point cloud data from airborne lidar;

[0047] The airborne lidar point cloud data includes the coordinates of scanning points on the bamboo surface and the laser reflection intensity, and records the average ground elevation; each scanning point on the bamboo surface has a unique laser reflection intensity.

[0048] In this embodiment, the airborne lidar point cloud data includes the coordinates of scanned points on the surface of the bamboo. and the laser reflection intensity at that point Where i is the sequential number of the scanning point coordinates on the bamboo surface, and the coordinates of all the scanning points on the bamboo surface constitute the scanning point set. The average ground elevation is... .

[0049] Step 2: Data preprocessing;

[0050] The collected point cloud data is preprocessed, including data noise reduction, ground threshold separation and normalization, to obtain the normalized result of laser reflection intensity.

[0051] Gaussian filtering was applied to the coordinates of all collected bamboo surface scanning points. Then, threshold separation and normalization were performed on the laser reflection intensity of each bamboo surface point. The specific process is as follows:

[0052] Coordinates of scanning points on the collected bamboo surface By using a preset formula:

[0053] ;

[0054] Gaussian filtering is performed to obtain the Gaussian filtered output of the coordinates of each scanning point on the bamboo surface. ,in σ represents the Gaussian weights; where σ is the preset standard deviation of the Gaussian filter, which controls the smoothness of the filter. By adjusting the standard deviation of the Gaussian filter, noise points can be effectively removed while retaining the main features of the point cloud. Its preset value is 0.2m.

[0055] Furthermore, ground points are separated by elevation thresholds to obtain the above-ground and underground parts, and the laser reflection intensity of the above-ground part is normalized.

[0056] Gaussian filter output for the coordinates of each scanning point on the bamboo surface Its z-axis component is greater than the average ground elevation. The part is recorded as the above-ground part. The z-axis component is less than or equal to the average ground elevation. The part is recorded as the underground part. .

[0057] For the above-ground part Obtain the Gaussian filtered output of the coordinates of each scanning point on the surface of the bamboo. original coordinates Corresponding laser reflection intensity And through a preset formula:

[0058] ;

[0059] Map it to a fixed range: [0, 100], and use it as the Gaussian filtered output of the coordinates of each scanning point on the bamboo surface. The corresponding laser reflection intensity normalization result Smax and Smin represent the laser reflection intensity, respectively. The maximum and minimum values ​​in the range.

[0060] Step 3: Fitting the bamboo trunk axis;

[0061] Based on the normalized processing results of laser reflection intensity, the missing area is filled by fitting the bamboo stem axis through hierarchical reconstruction and principal component analysis.

[0062] The system is divided into layers at preset height intervals, and point cloud data matrices for each height layer are generated based on the laser reflection intensity normalization results. PCA principal component analysis is performed on the point cloud data matrix results for each height layer to calculate the eigenvectors of the point cloud covariance matrix. Discrete bamboo stem segments are generated based on the eigenvectors of the covariance matrix, and the bamboo stem axis is fitted. Virtual cloud points with preset density are generated based on the bamboo stem axis to fill the missing areas at the base of the bamboo stem.

[0063] It's important to note that due to the working principle of lidar and the specific structure of bamboo forests, laser scanning often cannot fully penetrate bamboo leaves and canopies, resulting in significant gaps in the point cloud data of the bamboo trunk area. Lidar point clouds are typically concentrated on the outer contours of the bamboo (i.e., the branches and leaves), while the point cloud for the central structure (such as the trunk) is sparse or completely absent. This situation makes it difficult for traditional point cloud segmentation and feature extraction methods to accurately reconstruct the complete morphology of the bamboo trunk, thus affecting the accuracy of tasks such as disaster damage assessment and ecological restoration. Therefore, generating virtual point clouds to fill in these missing areas of the bamboo trunk becomes crucial to compensate for the missing regions in the point cloud data. By generating virtual points on the bamboo trunk, we can achieve a complete reconstruction of the bamboo trunk axis, thereby improving the integrity and accuracy of the entire point cloud data and providing more precise basic data for subsequent disaster damage identification and ecological analysis.

[0064] In this embodiment, layers are created at preset height intervals of 0.5m. The specific process is as follows:

[0065] Gaussian filter output to obtain the coordinates of all scanning points on the surface of bamboo In the middle, the z-axis component of the above-ground part The maximum value zmax and minimum value zmin are determined by a preset formula:

[0066] ;

[0067] Calculate the height range of each height level k. , where k is the height number and int is the integer operator;

[0068] The laser reflection intensities were normalized based on the height range of each altitude layer. Assign the laser reflection intensity to the corresponding height layer k, and ensure that the normalized results of all laser reflection intensities satisfy:

[0069] ;

[0070] Generate a point cloud data matrix with a height layer:

[0071] ;

[0072] The first three columns of the point cloud data matrix represent the x-axis, y-axis, and z-axis components of the coordinates corresponding to the Gaussian filtered output of all bamboo surface scanning point coordinates contained in height layer k; the fourth column of the point cloud data matrix represents the normalized result of the laser reflection intensity corresponding to each coordinate; and n is the total number of bamboo surface scanning point coordinates contained in height layer k.

[0073] Furthermore, principal component analysis is performed on the point cloud data matrix at each altitude level using a preset formula:

[0074] ;

[0075] Please see Figure 2 As shown, calculate the covariance matrix of each height layer k. The corresponding eigenvalues ​​and eigenvectors are solved; the eigenvector corresponding to the largest eigenvalue is selected as the axis direction of the bamboo stem, and the bamboo stem segments corresponding to each height layer k are obtained; the bamboo stem axis is fitted based on the bamboo stem segments of each height layer.

[0076] In this embodiment, the fitting process is as follows:

[0077] By detecting spatially adjacent segments through connected component analysis, the endpoint of the bamboo stem segment corresponding to the upper height layer k-1 and the starting point of the bamboo stem segment corresponding to the lower height layer k are obtained at the interface between each height layer k and k-1. Segments between upper and lower layers whose x-axis or y-axis components of the starting point and the endpoint are less than a preset threshold of 0.1m are marked as adjacent segments. The endpoint of the upper height layer bamboo stem segment and the starting point of the lower height layer bamboo stem segment with the smallest distance among the adjacent segments are located, and the midpoint between them is taken as the fitting point of the bamboo stem axis. The fitting points of all bamboo stem axes are connected to obtain the fitted bamboo stem axis.

[0078] Furthermore, virtual cloud points with a preset density of 200 points / meter are generated based on the fitted bamboo trunk axis.

[0079] Step 4: Segmentation feature fusion and image post-processing;

[0080] Post-processing based on the fitted bamboo trunk axis enhances the image representation containing disaster damage type features.

[0081] The specific process for matching disaster loss types is as follows:

[0082] Please see Figure 3 As shown, if the fitted bamboo trunk axis is found to have no axis division, and the angle between the highest layer in the fitted bamboo trunk axis and the ground xoy plane is less than the first preset threshold angle: 30°, then it is determined to be a bent bamboo.

[0083] If the fitted bamboo stem axis is found to have no axis branching and there is a missing axis, it is determined to be a broken bamboo.

[0084] If the fitted bamboo stem axis is found to have branching axis, and the angle between the branching axis is greater than the second preset threshold angle: 15°, then it is determined to be a split bamboo.

[0085] If it is detected that at the root position, i.e. at height layer k=1, the angle between the fitted bamboo trunk axis and the ground xoy plane is less than the third preset threshold angle: 60°, then it is determined to be a turned bamboo.

[0086] Furthermore, based on the disaster damage type, the feature images of the bamboo trunk and canopy point clouds at each layer are post-processed in parallel to enhance the image representation containing disaster damage type features and reconstruct the three-dimensional structure of the entire bamboo forest. The specific process is as follows:

[0087] Please see Figure 4 As shown, parallel image post-processing is performed for each height layer to generate a 3D structural feature map of the entire bamboo forest. For bent bamboo, the axis is extended; for broken bamboo, the fracture surface with curvature exceeding a preset threshold is located, extended along the normal direction, and random noise is added to increase the point cloud density of the fracture surface, thus reinforcing the jagged edges; for split bamboo, branching point clouds are generated at the bamboo stem forks to reconstruct the fragmented morphology of the bamboo stem fibers; for bamboo with turned stumps, a compensation point cloud with an inclination angle multiplied by a density coefficient of 1.3 is generated at the root.

[0088] It is important to note that accurately identifying different damage types within bamboo forests is crucial for the overall restoration and planning of the forest during disaster damage assessment and ecological restoration. Through segmentation feature fusion and image post-processing, we can not only accurately match the damage types of bamboo trunks and canopies, but also further improve the accuracy of the bamboo forest's 3D structure reconstruction by enhancing image representation. The purpose of image post-processing is to enhance the features of different damage types, allowing for a clearer and more accurate representation of the specific damage to bamboo during post-disaster assessment. This process, through parallel processing of image features from point clouds of bamboo trunks and canopies at each layer, strengthens the image representation containing damage type features, providing more refined data support for subsequent restoration work. This processing not only improves image detail but also better reflects damage forms such as bending, breakage, splitting, and overturning of bamboo trunks, thus providing a more reliable 3D structural model for bamboo forest restoration, ecological planning, and other management work.

[0089] Second embodiment:

[0090] Please see Figure 5 As shown, taking a sample plot of bamboo forest affected by snow disaster in a certain area as an example, the specific process for identifying snow disaster damage to bamboo forests is as follows:

[0091] Step 1: Input airborne lidar point cloud and ground survey data; in the preprocessing stage, Gaussian filtering is used for noise reduction, and ground points are separated by elevation threshold. The point clouds of bamboo trunks (0-10 meters) and healthy tree canopies (greater than 10 meters) are divided according to the height threshold.

[0092] Step 2: Implement a layered reconstruction strategy to address the sparse point cloud at the base of moso bamboo: First, slice the bamboo stem point cloud into layers at 0.5-meter height intervals. Within each layer, calculate the eigenvectors of the point cloud covariance matrix using Principal Component Analysis (PCA) and select vertical segments with a z-axis component greater than 0.9. Second, detect spatially adjacent segments using connected component analysis and connect discrete segments of the same bamboo stem by setting an Euclidean distance threshold of 0.3 meters. Finally, fit the bamboo stem axis based on the center point of the connected segments and generate virtual point clouds along the axis at 0.02-meter intervals to fill the missing area at the base. The virtual point density is set to 200 points / meter.

[0093] Step 3: Based on the preliminary damage type identified after bamboo trunk densification, perform customized operations: For bent bamboo, use axis extension technology to extend the crown axis by 0.5 meters along the bending direction of the bamboo trunk and increase the point cloud density to 1.5 times the original density; for trellis bamboo, generate contact point clouds within 0.3 meters below the lowest point of the crown, with a density of 200 points / square meter to simulate the crown's contact with the ground; for broken bamboo, locate the fracture surface with a curvature greater than 0.2, extend it 0.3 meters along the normal direction and add ±0.02 meters of random noise, increasing the density to 1.8 times to strengthen the serrated edges; for split bamboo, generate 20-40 degree bifurcation point clouds at the bamboo trunk bifurcation points to reconstruct the bamboo trunk fiber fragmentation morphology; for overturned bamboo, extend the crown by 0.8 meters along the inclined axis and generate a compensation point cloud with a density coefficient of 1.3 times according to the inclination angle to completely reconstruct the three-dimensional structure of the fallen bamboo.

[0094] Step 4: Parallel execution of bamboo trunk and canopy segmentation by the fusion segmentation module: Bamboo trunk segmentation adopts a dynamic density clustering algorithm, dynamically adjusting the neighborhood radius parameter according to the type of damage—for overturned bamboo, the neighborhood radius is set to 0.4 meters due to the tilted and dispersed bamboo trunk, and for broken bamboo, the radius is reduced to 0.3 meters due to the dense serrated features of the fracture surface area; Canopy segmentation calls a pre-trained PointNet++ model, inputting the four-dimensional features (three-dimensional coordinates + curvature + height + normal vector) of the encrypted point cloud, extracting damage-sensitive features through hierarchical feature learning and spatial transformation network, and outputting single-tree canopy segmentation results with probability labels;

[0095] Step 5: Tag fusion is mainly achieved through dual verification of spatial topological constraints (the Euclidean distance between the top of the bamboo trunk and the centroid of the tree crown is less than 1 meter) and disaster damage rules, ensuring that the bamboo trunk segmentation boundary and the tree crown segmentation form are strictly consistent in terms of disaster damage type logic. First, calculate the closest distance between the top of the bamboo trunk and the center of gravity of the tree crown. Then, verify the damage rules (specifically: for bent bamboo, the angle between the direction of the crown center of gravity offset and the direction of the bamboo trunk bending must be less than 30 degrees (calculate the angle between the center of gravity offset vector and the bending axis) to avoid contradictions between the crown offset direction and the bamboo trunk bending shape; for trellis bamboo, the crown height must be less than 1 meter (the z-value of the lowest point of the crown must be less than 1 meter) to ensure that the crown ground contact characteristics match the arched bamboo trunk; for broken bamboo, the crown isolation index must be greater than 0.5 (the minimum distance between the crown point cloud and the bamboo trunk exceeds 50% of the crown radius) to ensure that the crown suspension characteristics match the broken bamboo trunk; for bamboo with overturned stumps, the angle between the bamboo trunk axis and the main axis of the tree crown must be less than 10 degrees (calculate the angle between the two axes using principal component analysis) to ensure that the overall falling posture is consistent; for bamboo with broken bamboo, since it mainly shows local characteristics of the bamboo trunk and the crown shape does not change significantly, no mandatory verification rules are set). If a failure occurs, the matching is extended along the axis with the top of the bamboo trunk as the reference.

[0096] Step Six: After tag fusion is completed, the system verifies the segmentation results in three dimensions using ground survey data (damage types manually labeled by ground survey personnel): First, the accuracy rate of damage type identification is calculated by comparing the algorithm-labeled types such as bending, breakage, splitting, and uprooting with the field survey records for each bamboo stalk; second, the spatial location accuracy is evaluated through overlap rate analysis, calculating the distance error between the individual tree segmentation boundary and the measured bamboo stalk location; finally, damage feature morphology verification is performed by visualizing the encrypted point cloud using CloudCompare software and manually checking the consistency between key features such as the serrated structure of the fracture surface, the height of the canopy overhang, and the root exposure morphology with the actual damage morphology. After successful verification, the system outputs three main results: a spatial distribution map of individual tree damage types (including geometric boundaries and damage label vector data), a bamboo forest damage statistical report (statistical count / proportion / distribution density by type), and an encrypted point cloud dataset (LAS format file retaining damage features), supporting forestry departments to directly import the data into the GIS platform for disaster analysis and reconstruction planning.

[0097] It should be understood that the terms “comprising” and “including” used in this disclosure and claims indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0098] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the disclosure. As used in this disclosure and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this disclosure and claims means any combination and all possible combinations of one or more of the associated listed items, and includes such combinations;

[0099] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, characterized in that, Includes the following steps: Step 1: Data Acquisition and Input; Receives point cloud data from airborne lidar and ground survey data; Step 2: Data preprocessing; The collected point cloud data is preprocessed to obtain the normalized result of the laser reflection intensity; Step 3: Fitting the bamboo trunk axis; Based on the normalized processing results of laser reflection intensity, the missing area is filled by fitting the bamboo stem axis through hierarchical reconstruction and principal component analysis. Step 4: Segmentation feature fusion and image post-processing; Post-processing based on the fitted bamboo trunk axis enhances the image representation containing disaster damage type features. The specific process of post-processing based on the fitted bamboo stem axis is as follows: If it is found that the fitted bamboo trunk axis has no axis bifurcation, and the angle between the highest layer in the fitted bamboo trunk axis and the ground xoy plane is less than the first preset threshold angle, then it is determined to be a bent bamboo. If the fitted bamboo stem axis is found to have no axis branching and there is a missing axis, it is determined to be a broken bamboo. If the fitted bamboo stem axis is found to have axis branches, and the angle between the axis branches is greater than the second preset threshold angle, it is determined to be a split bamboo. If it is detected that the angle between the fitted bamboo trunk axis and the ground xoy plane is less than the third preset threshold angle at the root position, it is determined to be a turned bamboo. The specific process for enhancing the image representation containing disaster damage type features is as follows: The axis of the bent bamboo is extended; for the fracture surface of the broken bamboo with a curvature greater than a preset threshold, it is extended along the normal direction and random noise is added to increase the point cloud density of the fracture surface to strengthen the serrated edge; for the split bamboo, the branch point cloud is generated at the branching point of the bamboo trunk to reconstruct the shape of the broken bamboo fibers; for the turned bamboo, the compensation point cloud with the tilt angle multiplied by the preset density coefficient is generated at the root.

2. The method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, as described in claim 1, is characterized in that... The specific point cloud data of the airborne lidar is as follows: The data includes the coordinates of scanning points on the bamboo surface and the laser reflection intensity. Each scanning point on the bamboo surface has a unique laser reflection intensity corresponding to it. The ground survey data specifically includes the average ground elevation.

3. The method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, as described in claim 1, is characterized in that... The specific process of preprocessing point cloud data is as follows: Gaussian filtering was applied to the coordinates of all collected bamboo surface scanning points, and then threshold separation and normalization were performed on the laser reflection intensity of each bamboo surface point. For the coordinates of the scanned points on the bamboo surface collected, Gaussian filtering is used to calculate the Gaussian filter output of the coordinates of each scanned point on the bamboo surface.

4. The method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, as described in claim 1, is characterized in that... The specific process for preprocessing the collected point cloud data is as follows: The Gaussian filter output of the coordinates of each bamboo surface scanning point is separated by elevation threshold to obtain the above-ground part and the underground part. The laser reflection intensity of the above-ground part is normalized. The part whose z-axis component is greater than the average ground elevation is recorded as the above-ground part, and the part whose z-axis component is less than or equal to the average ground elevation is recorded as the underground part. For the above-ground portion, the laser reflection intensity corresponding to the original coordinate point of the Gaussian filter output of the scanning point coordinates of each bamboo surface is obtained, and then mapped to a fixed range to obtain the laser reflection intensity normalization result.

5. The method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, as described in claim 1, is characterized in that... The specific process of fitting the bamboo stem axis is as follows: The data is layered at preset height intervals, and point cloud data matrices for each height layer are generated based on the laser reflection intensity normalization result. Principal component analysis (PCA) is performed on the point cloud data matrices for each height layer to calculate the eigenvectors of the point cloud covariance matrix. Discrete bamboo stem segments are then generated based on the eigenvectors of the covariance matrix. The maximum value of the z-axis component of the ground part is obtained from the Gaussian filter output of all scanning points on the bamboo surface, and the height range of each height layer is calculated. The laser reflection intensity is normalized and assigned to the corresponding height layer according to the height range of each height layer. Principal component analysis was performed on the point cloud data matrix of each height layer to calculate the covariance matrix of each height layer and solve for its corresponding eigenvalues ​​and eigenvectors. The eigenvector corresponding to the largest eigenvalue was selected as the axis direction of the bamboo trunk to obtain the bamboo trunk segments corresponding to each height layer. The bamboo trunk segments were then fitted and connected.

6. The method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, as described in claim 5, is characterized in that... The specific process of fitting and connecting bamboo stem segments is as follows: By detecting spatially adjacent segments through connected component analysis, the endpoints of bamboo stem segments corresponding to the upper and lower height layers are obtained at the interfaces of each height layer. Segments between upper and lower layers whose x-axis or y-axis components of the starting and ending points are less than a preset threshold are marked as adjacent segments. The endpoints of bamboo stem segments in the upper and lower height layers with the smallest distance among the adjacent segments are located, and their midpoint is taken as the fitting point of the bamboo stem axis. The fitting points of all bamboo stem axes are connected to obtain the fitted bamboo stem axis. Virtual cloud points with a preset density are generated based on the fitted bamboo trunk axis.

7. The method for identifying snow disaster damage in bamboo forests using airborne lidar based on a differentiated point cloud encryption strategy, as described in claim 5, is characterized in that... The point cloud data matrix is ​​as follows: The first three columns of the point cloud data matrix represent the x-axis, y-axis, and z-axis components of the coordinates corresponding to the Gaussian filtered output of all bamboo surface scanning points contained in the height layer; the fourth column of the point cloud data matrix represents the normalized result of the laser reflection intensity corresponding to each coordinate.

Images