Objective Based on GF-7 stereo image and LiDAR data, Canopy Height Models (CHM) are used to extract the height of Tianshan spruce (Picea tianshanensis) single-tree height, and compare the performance of different data sources in the precision of single-tree height extraction. The purpose is to provide a technical reference for the extraction of single-tree height in mountainous natural forests.
Methods The Tianshan spruce in the internship forest of Xinjiang Agricultural University was used as the research object, RTK was used to locate each tree and measure the height of single trees, GF-7 remote sensing image and airborne LiDAR point cloud were used as the data sources, GF-7 and LiDAR digital surface model (DSM) was generated by the semiglobal matching (SGM) algorithm and cloth simulation filtering (CSF), the GF-7 DSM of was was calibrated using control points and epipolar images to generate the GF-7 digital elevation model (DEM).Three canopy height models (LiDAR-CHM, GF-LiDAR-CHM, GF-CHM) were thus constructed to derive tree heights.
Results The accuracy of extracting single-tree height based on LiDAR was higher, with an average accuracy of 91.46%, the average accuracy of GF-LiDAR was 83.02%, and the accuracy of extracting single-tree height of GF-7 was lower than the former, with an average accuracy of 80.13%.
Conclusion The fabric simulation filter classification by airborne LiDAR point cloud data can obtain high-precision DEM data, GF-7 can generate high-precision DSM by semi-global matching algorithm, and the DEM generated by kernel line image can be used as a method of tree height extraction for forest resources survey with large investigation range and limited cost.
DownLoad: