CN-121582780-B - Single wood detection method, system and computer equipment

CN121582780BCN 121582780 BCN121582780 BCN 121582780BCN-121582780-B

Abstract

The application relates to a single wood detection method, a system and computer equipment, wherein the method comprises the steps of obtaining point cloud data of a target forest area, constructing a digital surface model of the target forest area according to the point cloud data, obtaining gradient characteristic information of the target forest area by adopting a preset multi-order partial derivative calculation mode based on elevation information of the digital surface model, segmenting crown plaques according to the gradient characteristic information to obtain a plurality of segmented crown plaques, and carrying out cluster analysis on the plurality of segmented crown plaques based on a preset multi-feature similarity weight calculation method, and merging the crown plaques belonging to the same single wood to obtain a single wood detection result. The method provided by the application can accurately acquire the crown boundary and the crown point, and the over-segmented crown plaques are combined according to the similarity weight value of multiple features, so that the segmentation error is effectively avoided, and the single wood detection precision is improved.

Inventors

ZHANG XIAOLI
Lei Lingting
CHAI GUOQI
Yao Zongqi

Assignees

北京林业大学

Dates

Publication Date: 20260512
Application Date: 20251120

Claims (7)

1. A method of single wood detection comprising: Acquiring point cloud data of a target forest area, and constructing a digital surface model of the target forest area according to the point cloud data; Based on the elevation information of the digital surface model, a preset multi-order partial derivative calculation mode is adopted to obtain gradient characteristic information of a target forest area, and the method comprises the following steps: Performing first-order partial derivative calculation on the digital surface model based on the digital surface model to obtain one-step degree information of the digital surface model, wherein the one-step degree information comprises one-step degree vector and one-step degree vector; Performing second-order partial derivative calculation on the digital surface model based on a first-order partial derivative calculation result of the digital surface model to obtain second-order gradient information of the digital surface model, wherein the second-order gradient information comprises a second-order gradient vector and a second-order gradient vector; Obtaining gradient characteristic information of a target forest area based on the first-step gradient information and the second-step gradient information, wherein the gradient characteristic information comprises crown elevation change rate and crown surface convexity; according to the gradient characteristic information, performing over-segmentation on the crown plaque to obtain a plurality of over-segmented crown plaques; Based on a preset multi-feature similarity weight calculation method, performing cluster analysis on the plurality of divided crown plaques, merging crown plaques belonging to the same single tree to obtain a single tree detection result, wherein the method comprises the following steps: Acquiring characteristic information of the plurality of overclashed crown patches, wherein the characteristic information comprises spectral characteristics, texture characteristics and color space characteristics; constructing a weighted undirected graph based on the characteristic information of the crown plaques, wherein the vertex of the weighted undirected graph is the center point of the crown plaques after over-segmentation, and the similarity between the crown plaques after over-segmentation is used as the similarity weight value of the connected edges; calculating a similarity weight value between adjacent crown plaques based on a preset similarity weight function; clustering analysis is carried out on the crown plaques based on the weight values among the crown plaques, and the crown plaques with similarity weight values higher than a preset value are combined into crown plaques of the same single tree, so that a single tree detection result is obtained; The calculating the similarity weight value between adjacent crown plaques based on the preset similarity weight function comprises the following steps: Respectively obtaining a plurality of spectral features, a plurality of texture features and feature variables of a plurality of color space features corresponding to each pixel of the plurality of overclashed crown patches, and carrying out normalization calculation to obtain average values corresponding to the plurality of spectral features, the plurality of texture features and the plurality of color space feature variables of each crown patch, wherein the spectral features, the texture features and the color space features respectively comprise a plurality of feature variables; respectively calculating weight values of the plurality of characteristic variables between adjacent crown plaques according to a preset weight calculation formula; And respectively selecting the characteristic variable with the largest weight value in the spectral characteristic, the texture characteristic and the color space characteristic as an input variable, and inputting the input variable into a preset similarity weight function to obtain a similarity weight value between adjacent crown patches.
2. The method for detecting single wood according to claim 1, wherein the step information of the digital surface model is obtained by performing first-order partial derivative calculation on the digital surface model, and the step information comprises: calculating gradient vectors of the digital surface model in the x direction and the y direction to obtain a gradient vector: Wherein, the And Respectively digital surface models First partial derivatives in the x-direction and y-direction, A step vector for the digital surface model; Calculating a step vector size of the digital surface model according to the step vector: Wherein, the A step vector size; the calculating the second partial derivative of the digital surface model based on the first partial derivative calculation result of the digital surface model to obtain second order gradient information of the digital surface model includes: calculating a second-order gradient vector of the digital surface model according to the one-step gradient vector: Wherein, the And The second partial derivatives of the digital surface model in the x-direction and the y-direction respectively, A second order gradient vector for the digital surface model; calculating the second-order gradient vector magnitude of the digital surface model according to the second-order gradient vector: Wherein, the Is the second order gradient vector magnitude.
3. The method for detecting single wood according to claim 1, wherein the step of performing over-segmentation on the crown plaque according to the gradient feature information to obtain a plurality of over-segmented crown plaques comprises the steps of: According to the gradient characteristic information, obtaining a crown boundary and a plurality of tree top points in a target forest area; And performing over-segmentation on the target forest area based on the crown boundary and a plurality of tree top points to obtain a plurality of over-segmented crown plaques.
4. The method of claim 1, wherein the similarity weight function is: Wherein, the For the similarity weight value between adjacent patches i and j, And The average of the spectral features of the crown plaque i and the crown plaque j, And The average of the texture features of crown plaque i and crown plaque j respectively, And The average of the color space characteristics of crown patch i and crown patch j, respectively.
5. The method for detecting single wood according to claim 1, wherein the obtaining point cloud data of the target forest area, and constructing a digital surface model of the target forest area according to the point cloud data, comprises: Acquiring point cloud data of a target forest area based on unmanned aerial vehicle multi-angle measurement and laser radar scanning; performing data preprocessing on the point cloud data; and constructing a digital surface model of the target forest area based on the preprocessed point cloud data.
6. A single wood inspection system, comprising: The data acquisition module is used for acquiring point cloud data of a target forest area and constructing a digital surface model of the target forest area according to the point cloud data; The multi-order gradient calculation module is used for obtaining gradient characteristic information of a target forest area by adopting a preset multi-order partial derivative calculation mode based on the elevation information of the digital surface model, and comprises the following steps: Performing first-order partial derivative calculation on the digital surface model based on the digital surface model to obtain one-step degree information of the digital surface model, wherein the one-step degree information comprises one-step degree vector and one-step degree vector; Performing second-order partial derivative calculation on the digital surface model based on a first-order partial derivative calculation result of the digital surface model to obtain second-order gradient information of the digital surface model, wherein the second-order gradient information comprises a second-order gradient vector and a second-order gradient vector; Obtaining gradient characteristic information of a target forest area based on the first-step gradient information and the second-step gradient information, wherein the gradient characteristic information comprises crown elevation change rate and crown surface convexity; the over-segmentation module is used for over-segmenting the crown plaque according to the gradient characteristic information to obtain a plurality of over-segmented crown plaques; The single wood detection module is configured to perform cluster analysis on the plurality of excessively-divided crown plaques based on a preset multi-feature similarity weight calculation method, and combine crown plaques belonging to the same single wood to obtain a single wood detection result, where the single wood detection result includes: Acquiring characteristic information of the plurality of overclashed crown patches, wherein the characteristic information comprises spectral characteristics, texture characteristics and color space characteristics; constructing a weighted undirected graph based on the characteristic information of the crown plaques, wherein the vertex of the weighted undirected graph is the center point of the crown plaques after over-segmentation, and the similarity between the crown plaques after over-segmentation is used as the similarity weight value of the connected edges; calculating a similarity weight value between adjacent crown plaques based on a preset similarity weight function; clustering analysis is carried out on the crown plaques based on the weight values among the crown plaques, and the crown plaques with similarity weight values higher than a preset value are combined into crown plaques of the same single tree, so that a single tree detection result is obtained; The calculating the similarity weight value between adjacent crown plaques based on the preset similarity weight function comprises the following steps: Respectively obtaining a plurality of spectral features, a plurality of texture features and feature variables of a plurality of color space features corresponding to each pixel of the plurality of overclashed crown patches, and carrying out normalization calculation to obtain average values corresponding to the plurality of spectral features, the plurality of texture features and the plurality of color space feature variables of each crown patch, wherein the spectral features, the texture features and the color space features respectively comprise a plurality of feature variables; respectively calculating weight values of the plurality of characteristic variables between adjacent crown plaques according to a preset weight calculation formula; And respectively selecting the characteristic variable with the largest weight value in the spectral characteristic, the texture characteristic and the color space characteristic as an input variable, and inputting the input variable into a preset similarity weight function to obtain a similarity weight value between adjacent crown patches.
7. A computer device comprising a processor, a memory and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the log detection method according to any one of claims 1 to 5 when the computer program is executed.

Description

Single wood detection method, system and computer equipment Technical Field The invention relates to the technical field of forest resource investigation, in particular to a single wood detection method, a single wood detection system and computer equipment. Background Forest is taken as an important component of the land ecological system, and plays an important role in maintaining the balance of the ecological system, protecting the biodiversity, slowing down the global climate change and the like. The single wood is used as a minimum constituent unit of a forest, and the structural parameters of the single wood are efficiently and accurately extracted to be key factors required by accurate assessment of carbon reserves, ecological environment modeling research and the like, and are vital to forest resource investigation, forest resource quality improvement and synergy and ecological system function maintenance. In view of the problems of forest resource distribution fragmentation and wide coverage range in China, the traditional manual field investigation method is time-consuming, labor-consuming, high in cost and greatly influenced by subjective factors, such as large deviation in crown width and tree height measurement, and is difficult to meet the requirements of continuous large-range forest resource accurate and efficient investigation. The novel remote sensing technology, in particular the unmanned aerial vehicle remote sensing technology, forms a multi-platform, multi-angle and multi-mode stereoscopic observation system, has the advantages of large-scale, continuous space and dynamic monitoring, and effectively solves the problem that investigation staff are difficult to deeply operate under complex terrain conditions. More single-wood segmentation algorithms have been developed by researchers over the last decades. Wherein. The digital surface model/canopy height model is the primary data source for the mono segmentation. However, the existing single-wood segmentation algorithm based on the digital surface model (Digital surface model, DSM)/canopy height model has over-segmentation and under-segmentation phenomena at different degrees, and the segmentation result has larger uncertainty. To solve this problem, researchers have attempted to combine algorithms that are prone to over-segmentation and under-segmentation to take full advantage of the respective algorithms. For example, the watershed segmentation algorithm combines the Ncut algorithm of point cloud space information, the Ncut algorithm of a mean shift algorithm combined tree top point detection strategy and the Ncut algorithm of a mark control watershed algorithm combined hyperspectral information. The method can reduce the number of nodes and the time complexity in a forest environment with small topographic relief and low canopy density, improves the segmentation precision to a certain extent, and reduces the over-segmentation and under-segmentation phenomena. However, for highly closed forest stands with different crown shapes, under-segmentation is still difficult to completely avoid in the watershed segmentation algorithm result, which makes it difficult to maximize the effect of different algorithm combination strategies. Disclosure of Invention Based on the above, the invention aims to provide a single wood detection method, a system and computer equipment, which can accurately acquire the tree crown boundary and the tree top point, merge the over-segmented tree crown plaques according to the similarity weight value of multiple features, effectively avoid segmentation errors and improve single wood detection precision. In a first aspect, the present application provides a method for detecting a single wood, comprising: Acquiring point cloud data of a target forest area, and constructing a digital surface model of the target forest area according to the point cloud data; Based on the elevation information of the digital surface model, a preset multi-order partial derivative calculation mode is adopted to obtain gradient characteristic information of a target forest area; according to the gradient characteristic information, performing over-segmentation on the crown plaque to obtain a plurality of over-segmented crown plaques; and carrying out cluster analysis on the plurality of the over-divided crown plaques based on a preset multi-feature similarity weight calculation method, and merging crown plaques belonging to the same single tree to obtain a single tree detection result. In a second aspect, the present application provides a single wood detection system comprising: The data acquisition module is used for acquiring point cloud data of a target forest area and constructing a digital surface model of the target forest area according to the point cloud data; The multi-order gradient calculation module is used for obtaining gradient characteristic information of a target forest area by adopting a preset multi-order partial deriv