Extraction of individual tree heights using airborne LiDAR and optical images

Abstract

A method of obtaining tree height by integrating airborne LiDAR and optical data was proposed to improve the accuracy of tree height extraction by LiDAR. To remove the influence of ground fluctuation, the denoised and filtered point cloud data is first normalized by elevation, and then the tree height is retrieved using individual tree segmentation. Then, from the optical remote sensing image, the object-oriented multi-scale segmentation technique is used to extract the boundary of an individual tree crown, and the normalized point cloud data is used to segment the single tree, yielding the position and height information of the individual tree. The R² of the tree height derived from airborne LiDAR data and the measured value is 0.93, the RMSE is 1.72m, and the standard nRMSE is 5.84 %, according to the experimental results. With the measured value, RMSE=1.03m, and standard nRMSE= 3.49 %, the tree height retrieved by mixing point cloud and optical image is R²=0.95. The precision of individual tree height extraction from airborne LiDAR point cloud data can be greatly improved by combining optical images.