Crown volume algorithm system

Abstract

According to the method, new-generation information acquisition and extraction technologies such as the unmanned aerial vehicle, the laser radar and optical remote sensing are comprehensively applied, so that comprehensive and accurate perception of forest data is realized, the working quality is ensured, the working efficiency is greatly improved, the cost investment is reduced, traditional field manual field investigation is replaced by aerial remote sensing investigation, and the method is suitable for large-scale popularization and application. The accuracy, reliability and high efficiency of advanced equipment can be brought into full play, high-precision laser radar scanning is performed on the forest, point cloud data is acquired, processing is performed through a big data analysis system, a forest crown volume measurement and calculation model is established, digital and intelligent forest management is realized, and meanwhile, a scientific basis is provided for industry or local standard establishment. The volume of the tree crown is accurately measured and calculated, and the problems of errors, disputes and the like caused by manual estimation in the market transaction process of a forestry system can be effectively solved. Basic data can be provided for national carbon-to-peak carbon neutralization, and powerful support is provided for realizing green and low-carbon development in China.

Description

Technical Field

[0001] Airborne LiDAR and intelligent measurement technology are used in forest felling area survey and design to accurately calculate tree crown volume. Background Art

[0002] Forest felling area survey design is an important part of forest resource management. Traditional forest felling area survey design has high labor intensity in field survey, long time for manual acquisition of tree data measurement, large workload, low work efficiency, and the quality of survey is also affected by human factors. Moreover, the error rate of using sample plots to represent the overall quantity is high, and it is impossible to scientifically and accurately calculate the real tree crown volume data, which is not conducive to forestry production and management decisions. Summary of the invention

[0003] 1. Main technical routes (1) Starting from 2020, a survey of typical sample plots of mature eucalyptus forests was carried out in 13 state-owned forest farms in Guangxi Zhuang Autonomous Region. By combining data collected by forestry experts over many years with data on more than 300 sample plots, data on canopy density, number of trees, average crown width, diameter at breast height, and tree height were obtained.

[0004] (2) Use UAV-mounted laser radar and optical camera to collect forest resource data in typical sample plots of mature eucalyptus forests.

[0005] (3) Develop and construct a laser radar data and optical camera aerial photography data analysis and processing system, use the laser radar point cloud data and optical camera aerial photography data to identify and automatically segment the tree crowns of the tree layer, calculate the number of trees and the crown area and average crown diameter of each tree, and use the laser radar point cloud data to calculate the tree height of the sample tree and the average height of the trees in the sample plot.

[0006] (4) Using the field survey data of the sample plots, LiDAR data and optical camera aerial photography data, a model for calculating the number of trees, tree crown area, tree diameter at breast height, tree height and crown volume was established.

[0007] (5) Calculate the crown volume of each economic timber species according to the crown volume calculation model of sample trees of each diameter class.

[0008] (6) Develop and construct an integrated system for calculating tree crown volume. Based on the collected LiDAR point cloud data and optical camera aerial photography data of the eucalyptus forest felling area, the system can achieve the effect of "one-click output" and produce small-group data and summary data of each felling area in one click.

[0009] 2. Main operation process of remote sensing data acquisition and processing 1. Data acquisition process (1) The operation process is divided into project scope determination, project task assignment, organization of the combination of drone + lidar + optical camera equipment, route planning based on the point cloud density requirements of tree measurement, and data collection.

[0010] (2) After data collection, data solution and preprocessing are carried out to provide qualified point cloud data and multispectral image data.

[0011] (II) Data processing flow (1) After data acquisition, submit qualified point cloud data results to the office and conduct data analysis.

[0012] (2) Conduct internal segmentation through data analysis and check factors such as data accuracy and quality.

[0013] (3) Task distribution and data processing. Data processing includes data clipping, point cloud adjustment, denoising, classification, manual fine classification, vegetation classification and stratification, single tree segmentation, fine single tree identification, tree height, crown area and average diameter measurement, etc.

[0014] 3. Main technical indicators The main technical indicators for the eucalyptus forest felling area survey shall comply with the relevant provisions of the "Guangxi Zhuang Autonomous Region Felling Area Survey and Design Technical Regulations", namely: (1) Accuracy of area survey of small groups in felled areas: ≥95%; (2) Forest volume and economic timber output: ≥90%; (3) Survey accuracy of survey factors related to forest stock measurement: tree height ≥ 95%, tree number ≥ 95%, crown area and average diameter ≥ 90%, tree breast diameter ≥ 90%.

[0015] IV. Main Technical Achievements 1. Technical parameter system: (1) Basic climatic conditions, optimal routes, altitudes, speeds, etc. for UAV flights.

[0016] (2) Economically feasible lidar point cloud parameters.

[0017] (3) Optical image imaging parameters.

[0018] (B) Automatic tree height measurement system.

[0019] (III) System for automatic identification and segmentation of individual trees and their crowns, and calculation of the number of trees, crown area and average diameter.

[0020] 4. Forest crown volume measurement system.

[0021] 5. Development steps of crown volume algorithm system (see attached Figure 1 ) 6. Modeling Process 1. Data Collection ●Collect original laser point cloud data ● Manual measurement of breast diameter 2. ETL of raw data ETL stands for Extraction, Transformation, and Loading of data. Only after the original data is processed by ETL can it be formed into a regular format to facilitate subsequent calculations and modeling.

[0022] The first step of ETL is data extraction. This is the stage where the data pipeline extracts raw data from multiple data sources of data acquisition and consolidates it into a staging area.

[0023] The second step of ETL is to transform the data, converting the data in the staging area into the correct format to make it suitable for subsequent modeling calculations. In its original source, the data format is diverse and chaotic and cannot be used directly. In the transformation stage, the data will be structured and reformatted to make it suitable for subsequent modeling requirements.

[0024] The final step of ETL is to load the transformed and properly formatted data in the data warehouse. All the data can be loaded at once (full load) or at scheduled intervals (incremental load). This is achieved using batch or streaming loading. Batch loading is when the ETL program extracts batches of data from the source data based on a schedule (e.g., every hour). Streaming ETL (also known as real-time ETL or stream processing) is an alternative approach where the data pipeline brings in information as soon as the source data provides it.

[0025] Computation and modeling The calculation formula of DBH is as follows: d = c1 * (h+cw) ** c2 Among them, d is the breast diameter, c1 is the constant coefficient, h is the tree height, cw is the crown diameter, and c2 is the exponential constant.

[0026] The ETL process sorts out a sample data set from the original data. The instance of the data set contains the two features h and cw in the above formula and the d value as a label.

[0027] Use Python programming language to write a program, use numpy, scipy and other scientific computing modules to load the sample data set in the data warehouse into the memory in the form of vectors. Then use the curve fitting calculation method, set the initial value, substitute it into the above formula for a large number of iterations, and fit the c1 and c2 constants.

[0028] For unlabeled datasets, the iteratively calculated c1 and c2 are used as parameters, the known tree height and crown diameter are passed in, and the unknown breast diameter is calculated by referring to the above formula. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] one, Figure 1 : This flowchart intuitively reflects the 12 steps of crown volume algorithm system development.

[0030] two, Figure 2 : Use drone-mounted lidar + optical camera to collect forest resource data in typical sample plots of mature eucalyptus forests.

[0031] three, Figure 3 :LiDAR point cloud data and optical camera aerial photography data are imported into the system to identify and automatically segment the tree crown of the tree layer and calculate the number of trees.

[0032] Four, Figure 4 : Use lidar point cloud data to calculate the height, crown diameter, crown area and volume of each tree in each forest plot.

[0033] five, Figure 5 : Import lidar point cloud data and optical camera aerial data, and the system automatically identifies the coordinates of each tree.

[0034] six, Figure 6 :The system calculates all the parameters of the measured forest land, and finally calculates the crown volume of the forest land.

Claims

1. 1 Main technical routes (1) Starting from 2020, a field survey of typical sample plots of mature eucalyptus forests was carried out in thirteen state-owned forest farms in Guangxi. Through more than 300 sample plots combined with data collected by forestry experts over the years, data on the density, number of trees, average crown width, breast diameter, and tree height were obtained.

2. (2) Use UAV-mounted LiDAR and optical camera to collect forest resource data of typical mature eucalyptus forest plots.

3. (3) Develop and construct a system for analyzing and processing laser radar data and optical camera aerial photography data. Use laser radar point cloud data and optical camera aerial photography data to identify and automatically segment tree crowns, calculate the number of trees, the crown area of ​​each tree, and the average crown diameter. Use laser radar point cloud data to calculate the height of sample trees and the average height of trees in sample plots.

4. (4) Using the field survey data of the sample plots, LiDAR data and optical camera aerial photography data, a model for calculating the number of trees, tree crown area, tree diameter at breast height, tree height and crown volume was established.

5. (5) Calculate the crown volume of each economic timber species according to the crown volume calculation model for each diameter class of trees.

6. (6) Develop and construct an integrated system for calculating tree crown volume. Based on the collected LiDAR point cloud data and optical camera aerial photography data of the eucalyptus forest felling area, the system can achieve the effect of "one-click output" and produce small-group data and summary data of each felling area in one click.

Images