Semi-automatic segmentation of wood and foliage using Terrestrial Laser Scanner

Terrestrial Laser Scanner (TLS), based on Lidar technology, has been widely applied for a number of environmental applications. Several authors have reported the use of TLS for forestry applications and in forest inventory (i.e. tree height, diameter at breast height, tree distribution, etc.). TLS technology can be an effective alternative to overcome the limitations of the conventional ground based forest inventory techniques: expensive, time consuming, limited accuracy, destructive measurements. In addition, post-processing of TLS point clouds could enable extensive analysis of data by automatic or semi-automatic methods. Recent applications of TLS data analysis have been focused on detailed description of the canopy structure: canopy density, leaf area density, crown bulk density, etc. However, the operational use of TLS techniques for canopy characterization of broadleaf forests needs further investigations. In particular, segmentation between points representing woody material, leaves and small branches is a key factor to improve the accuracy of tree and canopy description. The main objective of this work was to develop a semi-automatic segmentation method of broadleaf tree species for improving the estimate of both canopy density distribution and woody material volumes. A voxel-based approach was developed and tested using a TLS data set collected in field by multiple scanning on four cork oak trees. After using noise reduction filters, voxels were used as input to generate clusters through a point density algorithm. Clustering process led to the identification of wood and leaf voxels. Points belonging to each voxel were then classified and quantified as wood, foliage and noise. Experimental results show that the semi-automatic segmentation algorithm can accurately discriminate wood and foliage clusters and consequently give the points of cloud associated to foliage, trunk and main branches.