Fusion of airborne LiDAR and digital photography data for tree crowns segmentation and measurement

Abstract

During airborne laser scanning, different types of information are available including Light Detection And Ranging (LiDAR) data as a cloud of 3D points, aerial digital photography data (hyperspectral or color visual images), and additional information about parameters of shooting. Difficulties of large image stitching due to parallax effects lead to distortions between ground truth 3D LiDAR coordinates and 2D visual coordinates of the same point. Our contribution is to develop a method for fusion of LiDAR and visual information for accurate segmentation of individual tree crowns in order to receive biomass measurements. The shearlet theory was used to improve boundaries and texture of airborne images. Also in this paper, a higher-order active contour model is applied for area evaluation of tree crowns in a plane. The received area measurements are promising and coincide with expert estimations providing accuracy 92–96%. The modeling results are good for non-Lambert space of forest.