92 / 2023-05-24 10:29:23

Trunk point cloud extraction of natural forest in mining area using multi-scale characteristics and random forest

Trunk point cloud extraction, natural forest, mining area, multi-scale characteristics, random forest

The vegetation of the mining area has a relatively strong carbon fixation rate and potential. It plays an energy dynamic role in the ecosystem and plays an increasingly important role, such as mitigating regional climate change, water conservation, water and soil conservation, adjusting carbon balance and many other aspects. Under the current "carbon neutrality" background, three-dimensional vegetation monitoring and restoration of the mining area have become an important part of mining restoration.

The continuous development of light detection and ranging (LiDAR) technology has provided new technological means for three-dimensional monitoring of vegetation in mining areas. In recent years, handheld LiDAR has been continuously developed, which not only has great portability but also is not limited by the site. It can enter almost all types of areas for scanning. In the field of forest sample survey, the application of handheld LiDAR is changing the three-dimensional information collection method of forest canopy structure three-dimensional monitoring, providing a new research perspective for the three-dimensional monitoring of vegetation in mining areas. Existing research is mostly based on data collection and extraction of artificial forests. Artificial forests have characteristics such as uniform vegetation distribution, orderly individual growth, and uniform and reasonable population structure. However, the vegetation in mining areas is mostly natural forests, and the growth process is often affected by the natural environment, the forests are of different ages and the structure is relatively complex, with enormous water conservation and natural disaster prevention and control capabilities. There are significant differences from artificial forests, with rich biodiversity, making existing trunk extraction methods less adaptable to natural forest trunks in complex mining environments.

 In order to quantitatively depict the three-dimensional structure information of natural forest trees in the complex environment of mining area, a method of extracting the trunk point cloud of natural forest trees in mining area using multi-scale features and random forest is proposed. On the basis of analyzing the elevation, echo, intensity and other attribute characteristics of point cloud data in the mining area, the multi-scale perpendicularity characteristics, normalized height, height statistics, surface characteristics, spatial distribution characteristics, echo characteristics, intensity characteristics and other feature parameters are extracted and constructed. The feature selection algorithm of random forest is used to optimize the classification feature set, and then the point cloud of the trunk of natural forest in mining area is extracted using random forest method. Vegetation from mining areas in the Henan section of the Yellow River Basin is selected for experiments, the experimental results show that the feature selection method based on random forest can effectively reduce the feature dimension, and compare and analyze the importance measure of features.