Forest crown disease and insect pest feature extraction method and system based on laser radar data

Abstract

The invention provides a canopy disease and insect pest feature extraction method and system based on laser radar data, and the method comprises the steps: carrying out the three-dimensional scanning of a target forest through a multi-echo laser radar, obtaining the multi-echo point cloud data of the target forest, carrying out the preprocessing of the multi-echo point cloud data, and obtaining the feature of the canopy disease and insect pest feature of the target forest; and based on multi-dimensional information such as a three-dimensional space structure, reflection intensity distribution and echo width characteristics of laser point cloud in the pure forest crown point cloud data, quantitatively evaluating the health state of the forest crown trees in the target forest from three aspects of a tree crown weakness characteristic value, a leaf falling ratio and a disease and pest stress evaluation value. And extracting a three-dimensional feature vector reflecting the occurrence of diseases and pests as a comprehensive feature vector representing the diseases and pests of the crown trees. According to the feature extraction scheme provided by the invention, automatic conversion from target forest radar data to forest crown pest and disease damage features can be completed, quantitative evaluation of pest and disease damage is realized, and an accurate data basis is provided for pest and disease damage control of forest trees.

Description

Technical Field

[0001] The present invention belongs to the field of forest pest monitoring, and in particular relates to a method and system for extracting forest canopy pest features based on laser radar data. Background Art

[0002] In current forest management, LiDAR technology is widely used in forest resource surveys and monitoring, mainly to obtain tree parameters such as three-dimensional structure, tree height, and crown diameter. However, related research focuses on the measurement of tree parameters and the estimation of forest resources. There is no effective technical solution for the extraction and evaluation of pest and disease characteristics. The existing forest pest and disease monitoring solutions have the following problems:

[0003] (1) It mainly focuses on the macroscopic health assessment of trees, but fails to fully obtain the richer reflection information in the lidar multi-echo data, such as echo intensity, waveform width and other characteristics, making it difficult to accurately capture the subtle characteristics of pests and diseases inside the canopy;

[0004] (2) They often focus on a single tree health indicator, such as tree height, trunk diameter or leaf loss, but fail to comprehensively assess tree health from multiple dimensions. They also lack quantitative assessments of the severity of pests and diseases, and are unable to provide sufficiently detailed analysis of forest health status.

[0005] (3) Existing lidar technology is mainly used to measure tree parameters. It lacks a special extraction algorithm for pest and disease characteristics, making it difficult to obtain relevant pest and disease information directly from lidar data.

[0006] Therefore, we need to develop a method and system for extracting forest crown pests and diseases based on lidar data, which can obtain internal information of the forest crown and realize automatic conversion of lidar data to forest crown pests and diseases characteristics. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for extracting features of canopy pests and diseases based on lidar data, so as to solve the problems mentioned in the above background technology that the existing forest pest and disease monitoring schemes are not accurate enough in evaluation, do not take into account comprehensive factors, and have poor targeting.

[0008] To achieve the above objectives, on the one hand, the present invention provides a method for extracting features of canopy pests and diseases based on lidar data, and the method is specifically as follows:

[0009] Performing a three-dimensional scan of the target forest by a multi-echo laser radar to obtain multi-echo point cloud data of the target forest, emitting near-infrared laser pulses and receiving flight time and intensity information of echoes reflected from the canopy of the target forest;

[0010] Preprocessing the multi-echo point cloud data, including denoising, classification and format conversion, to obtain pure canopy point cloud data;

[0011] Based on the multi-dimensional information of the laser point cloud in the pure canopy point cloud data, the health status of the canopy trees in the target forest is quantitatively evaluated from three aspects: the characteristic value of canopy weakness, the proportion of leaf shedding, and the evaluation value of pest and disease stress, and a three-dimensional feature vector reflecting the occurrence of pests and diseases is extracted, wherein the multi-dimensional information includes three-dimensional spatial structure, reflection intensity distribution, and echo width characteristics;

[0012] Based on the three-dimensional spatial structure, the crown weakness characteristic value is obtained; based on the reflection intensity distribution, the leaf shedding ratio is obtained; based on the echo width characteristics, the pest and disease stress assessment value is obtained;

[0013] Based on the crown weakness characteristic value, the leaf shedding ratio, and the pest and disease stress assessment value, the three-dimensional characteristic vector is obtained. , F=[L, R, S] The three components correspond to the crown weakness characteristic value L, the leaf shedding ratio , pest and disease stress assessment value .

[0014] Based on the above scheme, the preprocessing specifically includes:

[0015] The statistical filtering method is used to automatically identify and remove noise points based on the local features of the point cloud to obtain denoised point cloud data, which is recorded as denoised point cloud data.

[0016] Based on the echo characteristics of the multi-echo point cloud, by setting classification rules or clustering methods, the vegetation points and non-vegetation points in the denoised point cloud data are automatically distinguished, and the vegetation point subset is extracted as pure canopy point cloud data to characterize the canopy structure characteristics of the target forest.

[0017] Based on the above scheme, the point cloud data of a single tree A in the pure canopy point cloud data is recorded as , only includes the laser return points of the crown part, each laser return point contains the three-dimensional space coordinates And echo attributes, the three-dimensional feature vector represents the comprehensive severity of the tree's pests and diseases, which is a three-dimensional feature vector The three components correspond to the crown weakness characteristic value L, the proportion of leaf shedding, , pest and disease stress assessment value .

[0018] Based on the above scheme, the extraction of the three-dimensional feature vector reflecting the occurrence of pests and diseases is specifically as follows:

[0019] Regarding the The complete tree crown point cloud is divided into n layers of equal height in the vertical direction. Each layer represents the cross-section of the tree crown structure within a certain height range. exist The height range on the axis is recorded as , total number of layers Calculated by the following formula:

[0020]

[0021] Among them, the symbol Indicates rounding up. The interval for stratification; Divided into Sub-point cloud, denoted as ,in Corresponding to Layer height range z minutes + i - 1 dz , z minutes + i dz ] All points within

[0022] After the stratification process, Layered tree crown point cloud data , count each layer of point cloud The number of points , and the proportion of the points in this layer to the total points: ;

[0023] Based on the layered tree crown point cloud data, the tree crown weakness characteristic value L is extracted, specifically: The volume of the layer point cloud is , can be approximately calculated as a cylinder, ,in For the The average radius of the layer point cloud; define the Point cloud density of the layer , reflects the average number of points in a unit volume, and the density of each layer of point cloud is sequentially composed into a density distribution vector D=[ D 1 , D 2 , ..., D n ] ; Through the attenuation evaluation function Measuring the density distribution vector The abnormal degree of the crown decay characteristic value L is obtained;

[0024] Based on the layered tree crown point cloud data, extract the leaf shedding ratio , specifically including: based on the reflection intensity attribute of the layered point cloud, calculating the average intensity value of each layer in the layered crown point cloud data , and get the crown intensity distribution vector I=[ I 1 , I 2 , ..., I n ] ; Through the fall-off detection function Identify abnormal areas with significantly lower intensity and first calculate The mean and standard deviation , find all the average intensity values Satisfy less than The continuous layer is judged as the leaf falling area, where is a sensitivity parameter that controls the strictness of the judgment. The collection of point clouds in the leaf falling area Divide by the total number of tree crown point clouds N to get the leaf shedding ratio ;

[0025] Extract the pest and disease stress assessment value based on the layered tree crown point cloud data , specifically including: calculating the average echo width of the layered tree crown point cloud data layer by layer , get the crown echo width distribution W=[ W 1 , W 2 , ..., W n ] ; Calculate the echo width difference between adjacent layers , get the echo width difference vector D W=[ D W 1 , D W 2 , ..., D W n - 1 ] ; Design stress assessment function , measures the fluctuation abnormality of the echo width difference vector, and the function outputs the pest stress assessment value , mapped to the range from 0 to 1, The larger the value, the higher the degree of pest and disease stress.

[0026] On the other hand, the present invention also provides a system for extracting features of forest crown pests and diseases based on laser radar data, comprising:

[0027] A data acquisition module, used to perform a three-dimensional scan of a target forest by a multi-echo laser radar, obtain multi-echo point cloud data of the target forest, emit near-infrared laser pulses and receive flight time and intensity information of echoes reflected from the canopy in the target forest;

[0028] A data preprocessing module, used for preprocessing the multi-echo point cloud data, including denoising, classification and format conversion, to obtain pure canopy point cloud data;

[0029] A feature extraction module is used to quantitatively evaluate the health status of the canopy trees in the target forest from three aspects: canopy weakness characteristic value, leaf loss ratio, and pest and disease stress assessment value based on the multi-dimensional information of the laser point cloud in the pure canopy point cloud data, and extract a three-dimensional feature vector reflecting the occurrence of pests and diseases. The multi-dimensional information includes three-dimensional spatial structure, reflection intensity distribution, and echo width characteristics.

[0030] Based on the above solution, the data preprocessing module includes:

[0031] A denoising unit is used to automatically identify and remove noise points based on local features of the point cloud using a statistical filtering method to obtain denoised point cloud data, which is recorded as denoised point cloud data;

[0032] A classification unit, which is used to automatically distinguish vegetation points from non-vegetation points in the de-noised point cloud data based on the echo features of the multi-echo point cloud by setting classification rules or clustering methods, and extract the vegetation point subset as pure canopy point cloud data for characterizing the canopy structure characteristics of the target forest;

[0033] Based on the above solution, the feature extraction module includes:

[0034] The point cloud stratification unit is used to perform stratification processing on the tree crown point cloud, dividing the complete tree crown point cloud into n layers of equal height in the vertical direction, and obtaining the stratified tree crown point cloud data, each layer represents the cross section of the crown structure within a certain height range;

[0035] A frailty extraction unit, used for extracting a crown frailty characteristic value based on the layered crown point cloud data;

[0036] A leaf shedding ratio extraction unit, used for extracting the leaf shedding ratio based on the layered tree crown point cloud data;

[0037] A pest and disease assessment extraction unit, used for extracting a pest and disease stress assessment value based on the layered tree crown point cloud data;

[0038] The pest and disease feature extraction unit is used to integrate the features of the three dimensions of the crown weakness characteristic value, the leaf shedding ratio, and the pest and disease stress assessment value to obtain the crown pest and disease feature vector.

[0039] Compared with the prior art, the present invention has the following advantages and effects:

[0040] (1) Through the layered point cloud analysis method, the lidar data including echo attributes such as reflection intensity and echo width are processed. By analyzing the point cloud distribution characteristics inside the canopy layer by layer, the details of each layer in the forest can be extracted more accurately, and the changes in the canopy structure and the impact of pests and diseases can be fully described;

[0041] (2) The tree health status was quantitatively assessed from three aspects: the characteristic value of crown weakness, the proportion of leaf loss, and the assessment value of pest and disease stress, and a three-dimensional feature vector reflecting the occurrence of pests and diseases was extracted, which improved the accuracy and reliability of pest and disease monitoring;

[0042] (3) A forest pest and disease feature extraction algorithm for multi-echo lidar data was proposed. It can directly obtain relevant pest and disease information from lidar data, quantify the health status of trees, realize the automation of pest and disease monitoring, reduce labor costs, and improve the efficiency of pest and disease monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0044] Figure 1 This is a flow chart of a method for extracting features of forest crown pests and diseases based on laser radar data provided by an embodiment of the present invention;

[0045] Figure 2 It is a structural diagram of a forest crown pest and disease feature extraction system based on lidar data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to more clearly explain the purpose, technical solutions and advantages of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The example implementation methods can be implemented in various forms and should not be understood as being limited to the examples described herein. On the contrary, these implementation methods are provided to make the present invention more comprehensive and complete, and to fully convey the concepts of the example implementation methods to those skilled in the art.

[0047] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present invention. However, those skilled in the art will appreciate that the technical solution of the present invention can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring various aspects of the present invention.

[0048] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0049] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0050] The present invention is described in detail below in conjunction with specific embodiments:

[0051] Example 1

[0052] As attached Figure 1 As shown, Embodiment 1 of the present invention provides a method for extracting features of canopy pests and diseases based on laser radar data, and the specific steps of the method are as follows:

[0053] Step S1, performing a three-dimensional scan of a target forest by a multi-echo laser radar to obtain multi-echo point cloud data of the target forest, emitting near-infrared laser pulses and receiving flight time and intensity information of echoes reflected from the canopy of the target forest;

[0054] Specifically, the three-dimensional spatial coordinates (x, y, z) of the laser point are calculated based on the flight time, and the vegetation characteristics are analyzed based on the intensity information;

[0055] Step S2, preprocessing the multi-echo point cloud data, including denoising, classification and format conversion, to obtain pure canopy point cloud data;

[0056] Step S3, based on the multi-dimensional information of the laser point cloud in the pure canopy point cloud data, quantitatively evaluate the health status of the canopy trees in the target forest from three aspects: canopy weakness characteristic value, leaf shedding ratio, and pest and disease stress assessment value, and extract a three-dimensional feature vector reflecting the occurrence of pests and diseases, wherein the multi-dimensional information includes three-dimensional spatial structure, reflection intensity distribution, and echo width characteristics;

[0057] Specifically, based on the three-dimensional spatial structure, the crown weakness characteristic value is obtained; based on the reflection intensity distribution, the leaf shedding ratio is obtained; based on the echo width feature, the pest and disease stress assessment value is obtained;

[0058] Based on the crown weakness characteristic value, the leaf shedding ratio, and the pest and disease stress assessment value, the three-dimensional characteristic vector is obtained. , F=[L, R, S] The three components correspond to the crown weakness characteristic value L, the leaf shedding ratio , pest and disease stress assessment value .

[0059] Preferably, in step S1, the infrared laser pulse selects a near-infrared band as the working wavelength of the laser radar, and the near-infrared band includes 1064nm;

[0060] It should be noted that the near-infrared band has a higher reflectivity on the vegetation surface, which is conducive to obtaining stable and clear canopy echo signals.

[0061] Preferably, in step S1, the main parameters of the multi-echo laser radar include: wavelength, transmission power, pulse frequency, scanning angle, and point cloud density;

[0062] Specifically, the main parameters can be optimized according to the crown characteristics of the target forest to improve the penetration ability of the multi-echo lidar of complex crowns; by collecting the main parameters, the lidar system can obtain high-quality multi-echo point cloud data that can reflect the three-dimensional structural characteristics of the crown in the target forest, namely, the multi-echo point cloud data.

[0063] Preferably, step S2 specifically includes:

[0064] Step S201, the denoising: using a statistical filtering method, automatically identifying and removing noise points based on local point cloud features such as point cloud density and average distance, specifically: determining the neighborhood of each point in the multi-echo point cloud data, and calculating the statistical indicators of the points in the neighborhood, marking the points that deviate from the normal range as noise and removing them, to obtain denoised point cloud data, recorded as denoised point cloud data.

[0065] Step S202, the classification: based on the echo characteristics of the multi-echo point cloud such as intensity and width, by setting classification rules or clustering methods, the vegetation points and non-vegetation points in the de-noised point cloud data are automatically distinguished, and the vegetation point subset is extracted as pure canopy point cloud data for characterizing the canopy structure characteristics of the target forest;

[0066] Step S203, the format conversion: converting the pure canopy point cloud data into a common standard format such as LAS, storing point attributes in the form of point records to facilitate data exchange and software processing, wherein the point attributes include coordinates, reflection intensity, echo number, and classification code.

[0067] Preferably, in step S3, taking the point cloud data of a single tree A in the pure canopy point cloud data as an example, it is recorded as , only includes the laser return points of the crown part, each laser return point contains the three-dimensional space coordinates and echo properties such as reflection intensity , echo width etc., the three-dimensional feature vector is a three-dimensional feature vector , characterizing the comprehensive severity of the tree's pests and diseases, the three-dimensional feature vector The three components correspond to the crown weakness characteristic value L, the proportion of leaf shedding , pest and disease stress assessment value .

[0068] Preferably, step S3 specifically includes:

[0069] Step S301: The complete crown point cloud is divided into n layers of equal height in the vertical direction. Each layer represents the cross-section of the crown structure within a certain height range. The point cloud distribution characteristics inside the crown are analyzed layer by layer to comprehensively describe the changes in crown structure and the impact of pests and diseases. exist The height range on the axis is recorded as , total number of layers Calculated by the following formula:

[0070]

[0071] Among them, the symbol Indicates rounding up. is the interval of the layers, which is a preset parameter and can be flexibly set according to factors such as tree species, tree height, point cloud density, etc. In this embodiment, it is preferably to between; Divided into Sub-point cloud, denoted as ,in Corresponding to Layer height range z minutes + i - 1 dz , z minutes + i dz ] All points within

[0072] After the stratification process, Layered tree crown point cloud data , count each layer of point cloud The number of points , and the proportion of the points in this layer to the total points: ; Number of layered point clouds and proportion It reflects the distribution characteristics of the vertical structure of the tree crown and provides data support for subsequent weakness analysis.

[0073] Step S302: extracting the crown weakness characteristic value L based on the layered crown point cloud data, specifically: The changes in point cloud density at different layers reflect the density differences in the internal structure of the tree crown, which can be used to assess the health of the tree. The volume of the layer point cloud is , can be approximately calculated as a cylinder, ,in For the The average radius of the layer point cloud; define the Point cloud density of the layer , reflects the average number of points in a unit volume, and the density of each layer of point cloud is sequentially composed into a density distribution vector D=[ D 1 , D 2 , ..., D n ] ; By designing a frailty assessment function Measuring the density distribution vector The abnormal degree of the crown decay characteristic value L is obtained;

[0074] It should be noted that crown decay will lead to a decrease in canopy density, and abnormal changes in density distribution are a quantitative expression of the degree of decay; the decay evaluation function The crown attenuation characteristic value L is obtained by calculating the difference between the actual density distribution and the standard density distribution of healthy crowns. The larger it is, the greater the debilitation;

[0075] Specifically, in this embodiment, by Divergence is the difference between the actual density distribution and the standard density distribution of healthy crowns. ,in is a density distribution template obtained based on a large number of healthy tree samples. The frailty evaluation function Output value That is, the crown weakness characteristic value L, the crown weakness characteristic value L is normalized and mapped to the range of 0 to 1. The larger the value, the more serious the degree of weakness.

[0076] Step S303: extract the leaf shedding ratio based on the layered tree crown point cloud data , specifically including: based on the reflection intensity attribute of the layered point cloud, calculating the average intensity value of each layer in the layered crown point cloud data , and get the crown intensity distribution vector I=[ I 1 , I 2 , ..., I n ] ; By designing the fall-off detection function Identify abnormal areas with significantly lower intensity and first calculate The mean and standard deviation , find all the average intensity values Satisfy less than The continuous layer is judged as the leaf falling area, where is a sensitivity parameter, which controls the strictness of the judgment. In this embodiment, it is preferably selected between 1.5 and 3. The set of point clouds of the leaf falling area Divide by the total number of tree crown point clouds N to get the leaf shedding ratio ;

[0077] For example, assume , if the average strength of three consecutive layers is lower than , then the crown area corresponding to the three layers of point cloud is judged as the leaf falling area, and the leaf falling detection function The output result is a collection of point clouds of the leaf-falling area. , and add its points to the total points of the crown Divide by, and you get the leaf shedding ratio , which reflects the percentage of the crown shedding area to the entire canopy and is an important indicator for quantifying the scope of pests and diseases;

[0078] It should be noted that the falling of leaves will cause the reflection intensity of the local point cloud of the tree canopy to decrease significantly. Therefore, in this embodiment, the proportion of falling leaves is obtained based on the reflection intensity. .

[0079] Step S304: extracting the pest and disease stress assessment value based on the layered tree crown point cloud data , specifically including: calculating the average echo width of the layered tree crown point cloud data layer by layer , get the crown echo width distribution W=[ W 1 , W 2 , ..., W n ] ; Calculate the echo width difference between adjacent layers , get the echo width difference vector D W=[ D W 1 , D W 2 , ..., D W n - 1 ] ; Design stress assessment function , measures the fluctuation abnormality of the echo width difference vector, and the function outputs the pest stress assessment value , mapped to the range from 0 to 1, The larger the value, the higher the degree of pest and disease stress;

[0080] Exemplarily, the L2 norm of the echo width difference vector is taken as As a measure of abnormality, the calculation formula is: The larger the norm value, the more violent the width fluctuation, which means the more serious the pest and disease stress is. Proportional to the L2 norm value, ,in is a normalization constant, which is taken as the maximum L2 norm of all tree samples;

[0081] It should be noted that pest and disease stress can cause damage to the tree canopy structure and cause abnormal fluctuations in the echo width. The echo width reflects the scattering and attenuation of the laser pulse in the canopy and can be used to evaluate the integrity of the canopy. The drastic fluctuations in the echo width difference indicate the mutation and fragmentation of the canopy structure, reflecting the severity of the pest and disease stress. Therefore, in this embodiment, the degree of pest and disease stress in the tree canopy is determined based on the echo width.

[0082] Step S305: Comprehensively calculate the crown weakness characteristic value L and the leaf shedding ratio , the pest and disease stress assessment value The three-dimensional features are used to obtain the pest and disease feature vector of the crown of a single tree A, that is, the three-dimensional feature vector , F=[L, R, S] ;

[0083] It should be noted that the pest feature vector It comprehensively depicts the three aspects of crown weakness, shedding, and stress, and is a quantitative representation of the extent of tree diseases and insect pests;

[0084] Specifically, in order to eliminate the influence of different dimensions, the pest and disease characteristic vector is normalized by the maximum and minimum value normalization method. Normalize the value range of each component to [0, 1] In the interval, the normalized eigenvector It facilitates horizontal comparison between different trees and provides a unified scale for subsequent tree health grading and regional pest and disease assessment.

[0085] Step S306: The above steps are for the point cloud of a single tree A. Perform feature extraction and repeat the above process for each tree in the target forest area, and finally obtain a complete pest and disease feature dataset for the entire area , which contains the normalized feature vector of each tree, the pest and disease feature dataset It can be used for subsequent tasks such as regional health assessment, pest and disease distribution mapping, and optimization of prevention and control measures.

[0086] It should be noted that, in this embodiment, the assumptions and empirical parameters involved in the algorithm, such as the horizontal cross-sectional shape of the crown, the stratification interval, the abnormal threshold, etc., may vary due to factors such as tree species, forest type, and site conditions. In practical applications, it is necessary to adjust and optimize the details of the algorithm appropriately according to the characteristics of the specific monitoring object, combined with expert knowledge and field investigations, to improve the reliability and effectiveness of feature extraction; in addition, it is necessary to regularly use ground survey data to verify and calibrate the results of feature extraction, and update and improve the monitoring system based on new observation data to ensure the accuracy and timeliness of the monitoring system.

[0087] In this embodiment, a three-dimensional scan of the target forest is performed by a multi-echo laser radar to obtain multi-echo point cloud data of the target forest, and the multi-echo point cloud data is preprocessed. Based on the three-dimensional spatial structure, reflection intensity distribution, echo width characteristics and other multi-dimensional information of the laser point cloud in the pure canopy point cloud data, the health status of the canopy trees in the target forest is quantitatively evaluated from three aspects: crown decay characteristic value, leaf shedding ratio, and pest and disease stress assessment value, and a three-dimensional feature vector reflecting the occurrence of pests and diseases is extracted as a comprehensive feature vector of pests and diseases that characterizes the canopy trees. A three-dimensional feature vector that comprehensively reflects the severity of individual tree pests and diseases is constructed from the three aspects of crown decay, shedding, and stress, which provides a new idea for quantitative evaluation, realizes the automation and scale of laser radar forest pest and disease monitoring, improves monitoring efficiency and reduces costs.

[0088] Example 2

[0089] As attached Figure 2 As shown, Embodiment 2 of the present invention provides a system 200 for extracting features of canopy pests and diseases based on laser radar data. The execution of the system module includes:

[0090] The data acquisition module 201 performs a three-dimensional scan of the target forest by a multi-echo laser radar to obtain multi-echo point cloud data of the target forest, emits near-infrared laser pulses and receives flight time and intensity information of the reflected echo from the canopy of the target forest;

[0091] The data preprocessing module 202 preprocesses the multi-echo point cloud data, including denoising, classification and format conversion, to obtain pure canopy point cloud data;

[0092] Feature extraction module 203, based on the multi-dimensional information of the laser point cloud in the pure canopy point cloud data, quantitatively evaluates the health status of the canopy trees in the target forest from three aspects: canopy weakness characteristic value, leaf loss ratio, and pest and disease stress assessment value, and extracts a three-dimensional feature vector reflecting the occurrence of pests and diseases. The multi-dimensional information includes three-dimensional spatial structure, reflection intensity distribution, and echo width characteristics.

[0093] Preferably, the data preprocessing module 202 includes a denoising unit 2021, a classification unit 2022, and a format conversion unit 2023, which are specifically implemented as follows:

[0094] The denoising unit 2021 adopts a statistical filtering method to automatically identify and remove noise points based on local features of the point cloud, thereby obtaining denoised point cloud data, which is recorded as denoised point cloud data;

[0095] The classification unit 2022, based on the echo features of the multi-echo point cloud, realizes automatic distinction between vegetation points and non-vegetation points in the de-noised point cloud data by setting classification rules or clustering methods, and extracts the vegetation point subset as pure canopy point cloud data for characterizing the canopy structure characteristics of the target forest;

[0096] The format conversion unit 2023 is used to convert the pure canopy point cloud data into a universal standard format, and store the attributes of the points in the form of point records to facilitate data exchange and software processing. The attributes of the points include coordinates, reflection intensity, number of echoes, and classification codes.

[0097] Preferably, the feature extraction module 203 includes a point cloud stratification unit 2031, a weakness extraction unit 2032, a shedding ratio extraction unit 2033, a pest assessment extraction unit 2034, and a pest feature extraction unit 2035, which are specifically implemented as follows:

[0098] The point cloud stratification unit 2031 is used to perform stratification processing on the tree crown point cloud, dividing the complete tree crown point cloud into n layers of equal height in the vertical direction, and obtaining stratified tree crown point cloud data, each layer represents the cross section of the crown structure within a certain height range;

[0099] A frailty extraction unit 2032 is used to extract a crown frailty characteristic value based on the layered crown point cloud data;

[0100] A leaf shedding ratio extraction unit 2033 is used to extract a leaf shedding ratio based on the layered tree crown point cloud data;

[0101] A shedding ratio extraction unit 2034 is used to extract a pest and disease stress assessment value based on the layered crown point cloud data;

[0102] The pest and disease feature extraction unit 2035 is used to integrate the features of the three dimensions of the crown weakness characteristic value, the leaf shedding ratio, and the pest and disease stress assessment value to obtain a crown pest and disease feature vector.

[0103] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present invention, which follows the general principles of the present invention and includes common knowledge or customary techniques in the art that are not disclosed by the present invention. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present invention are indicated by the claims. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for extracting features of forest crown pests and diseases based on LiDAR data, characterized in that: include: Performing a three-dimensional scan of the target forest by a multi-echo laser radar to obtain multi-echo point cloud data of the target forest, emitting near-infrared laser pulses and receiving flight time and intensity information of echoes reflected from the canopy of the target forest; Preprocessing the multi-echo point cloud data, including denoising, classification and format conversion, to obtain pure canopy point cloud data; Based on the multi-dimensional information of the laser point cloud in the pure canopy point cloud data, the health status of the canopy trees in the target forest is quantitatively evaluated from three aspects: the characteristic value of canopy weakness, the proportion of leaf shedding, and the evaluation value of pest and disease stress, and a three-dimensional feature vector reflecting the occurrence of pests and diseases is extracted, wherein the multi-dimensional information includes three-dimensional spatial structure, reflection intensity distribution, and echo width characteristics; Based on the three-dimensional spatial structure, obtaining the crown weakness characteristic value; Based on the reflection intensity distribution, obtaining the leaf shedding ratio; Based on the echo width feature, obtaining the pest and disease stress assessment value; Based on the crown weakness characteristic value, the leaf shedding ratio, and the pest and disease stress assessment value, the three-dimensional characteristic vector is obtained. , The three components correspond to the crown weakness characteristic value L, the leaf shedding ratio , pest and disease stress assessment value .

2. The method for extracting forest crown pests and diseases features based on lidar data according to claim 1, characterized in that: The preprocessing specifically includes: The statistical filtering method is used to automatically identify and remove noise points based on the local features of the point cloud to obtain denoised point cloud data, which is recorded as denoised point cloud data. Based on the echo characteristics of the multi-echo point cloud, by setting classification rules or clustering methods, the vegetation points and non-vegetation points in the denoised point cloud data are automatically distinguished, and the vegetation point subset is extracted as pure canopy point cloud data to characterize the canopy structure characteristics of the target forest.

3. The method for extracting forest crown pests and diseases features based on LiDAR data according to claim 2, characterized in that: The point cloud data of a single tree A in the pure canopy point cloud data is recorded as , only includes the laser return points of the crown part, each laser return point contains the three-dimensional space coordinates and echo properties.

4. The method for extracting features of forest crown pests and diseases based on lidar data according to claim 3, characterized in that: right The complete tree crown point cloud is divided into n layers of equal height in the vertical direction. Each layer represents the cross-section of the tree crown structure within a certain height range. exist The height range on the axis is recorded as , total number of layers Calculated by the following formula: , where the symbol Indicates rounding up. The interval for stratification; Divided into Sub-point cloud, denoted as ,in Corresponding to Layer height range All points within After the stratification process, Layered tree crown point cloud data , count each layer of point cloud The number of points , and the proportion of the points in this layer to the total points: .

5. The method for extracting features of forest crown pests and diseases based on lidar data according to claim 4, characterized in that: Based on the layered tree crown point cloud data, the tree crown weakness characteristic value L is extracted, specifically: The volume of the layer point cloud is , can be approximately calculated as a cylinder, ,in For the The average radius of the layer point cloud; define the Point cloud density of the layer , reflects the average number of points in a unit volume, and the density of each layer of point cloud is sequentially composed into a density distribution vector ; Through the attenuation evaluation function Measuring the density distribution vector The abnormal degree of the crown decay characteristic value L is obtained.

6. The method for extracting features of forest crown pests and diseases based on lidar data according to claim 4, characterized in that: Based on the layered tree crown point cloud data, extract the leaf shedding ratio , specifically including: based on the reflection intensity attribute of the layered point cloud, calculating the average intensity value of each layer in the layered crown point cloud data , and get the crown intensity distribution vector ; Through the fall-off detection function Identify abnormal areas with significantly lower intensity and first calculate The mean and standard deviation , find all the average intensity values Satisfy less than The continuous layer is judged as the leaf falling area, where is a sensitivity parameter that controls the strictness of the judgment. The collection of point clouds in the leaf falling area Divide by the total number of tree crown point clouds N to get the leaf shedding ratio .

7. The method for extracting forest crown pests and diseases features based on LiDAR data according to claim 4, characterized in that: Extract the pest and disease stress assessment value based on the layered tree crown point cloud data , specifically including: calculating the average echo width of the layered tree crown point cloud data layer by layer , get the crown echo width distribution ; Calculate the echo width difference between adjacent layers , get the echo width difference vector ; Design stress assessment function , measures the fluctuation abnormality of the echo width difference vector, and the function outputs the pest stress assessment value , mapped to the range from 0 to 1, The larger the value, the higher the degree of pest and disease stress.

8. A forest crown pest feature extraction system based on LiDAR data, characterized in that: include: A data acquisition module, used to perform a three-dimensional scan of a target forest by a multi-echo laser radar, obtain multi-echo point cloud data of the target forest, emit near-infrared laser pulses and receive flight time and intensity information of echoes reflected from the canopy in the target forest; A data preprocessing module, used for preprocessing the multi-echo point cloud data, including denoising, classification and format conversion, to obtain pure canopy point cloud data; A feature extraction module is used to quantitatively evaluate the health status of the canopy trees in the target forest from three aspects: canopy weakness characteristic value, leaf loss ratio, and pest and disease stress assessment value based on the multi-dimensional information of the laser point cloud in the pure canopy point cloud data, and extract a three-dimensional feature vector reflecting the occurrence of pests and diseases. The multi-dimensional information includes three-dimensional spatial structure, reflection intensity distribution, and echo width characteristics.

9. The system for extracting forest crown pests and diseases features based on lidar data according to claim 8, characterized in that: The data preprocessing module comprises: A denoising unit is used to automatically identify and remove noise points based on local features of the point cloud using a statistical filtering method to obtain denoised point cloud data, which is recorded as denoised point cloud data; A classification unit, which is used to automatically distinguish vegetation points from non-vegetation points in the de-noised point cloud data based on the echo features of the multi-echo point cloud by setting classification rules or clustering methods, and extract the vegetation point subset as pure canopy point cloud data for characterizing the canopy structure characteristics of the target forest; The feature extraction module comprises: The point cloud stratification unit is used to perform stratification processing on the tree crown point cloud, dividing the complete tree crown point cloud into n layers of equal height in the vertical direction, and obtaining the stratified tree crown point cloud data, each layer represents the cross section of the crown structure within a certain height range; A frailty extraction unit, used for extracting the crown frailty characteristic value based on the layered crown point cloud data; A leaf shedding ratio extraction unit, used for extracting the leaf shedding ratio based on the layered tree crown point cloud data; A pest and disease assessment extraction unit, used for extracting the pest and disease stress assessment value based on the layered tree crown point cloud data; The pest and disease feature extraction unit is used to integrate the features of the three dimensions of the crown weakness characteristic value, the leaf shedding ratio, and the pest and disease stress assessment value to obtain the crown pest and disease feature vector.

Images