A Single Tree Recognition Method Based on Integrated Features of Lidar Waveform

Abstract

The invention discloses a light detection and ranging (LiDAR) waveform comprehensive feature-based individual tree identification method. The method includes the following steps of: acquiring data by means of an airborne small-light spot full-waveform LiDAR sensor; preprocessing LiDAR waveform data: positioning individual trees and extracting canopy spread; correcting the LiDAR waveform data based on emitted energy and distance information of the sensor and ground objects; constructing a voxel framework and performing structural decomposition of the LiDAR waveform data; extracting waveform feature variables of the individual trees; extracting point cloud feature variables of the individual trees; and screening out optimal feature variables by using a random forest method, and classifying tree species. As indicated by verification results, compared with others method in which remote sensing is adopted, the overall accuracy of the method is improved by about 15%, and the Kappa coefficient of the method is improved by about 0.13.

Description

technical field [0001] The invention relates to the technical field of forest resources management and protection, in particular to a single tree identification method based on LiDAR waveform comprehensive features. Background technique [0002] Accurate tree species classification is of great significance for forest resources survey, dynamic monitoring and biodiversity research, as well as for simulating single tree growth of specified tree species. At the same time, the information can also be used for forest resources survey, dynamic monitoring and biodiversity research, thus providing accurate data support for small-scale and medium-scale forest resource planning and intensive management. Conventional forest tree species survey methods mainly rely on field surveys and interpretation of large-scale aerial photographs, etc., and their accuracy is often low, and it is difficult to be practically promoted in large areas. LiDAR (Light Detection and Ranging) is an active remo...

AI Technical Summary

Problems solved by technology

However, most of the above methods are suitable for forest classification research with relatively simple tree species composition, and the classification accuracy is not high in forests with complex stand composition and structure

And only from a single point of view to mine LiDAR data, that is, the three-dimensional spatial information contained in the "point cloud" data and the geometric and energy information contained in the "waveform" data are not combined, and the potential of LiDAR data cannot be fully exploited.

Images