Automated detection of 3D individual trees along urban road corridors by mobile laser scanning systems

300 words): Recently, mobile laser scanning (MLS) has emerged as an efficient means to acquire massive 3D point clouds along urban road corridors for the several applications such as building footprint reconstruction, facade modeling and road traffic inventories. In this paper, we propose an automated strategy to address the issue of detecting 3D individual trees in urban traffic corridors using MLS data. Firstly, an approach based on analyzing the features in the spatial accumulation map of MLS point cloud is developed to identify manmade objects from natural objects by sequentially extracting man-made structures. The remaining points of natural objects should mainly contain vegetation and few vertical disturbing objects. This point class is further delivered to the next step to undergo a 3D segmentation process to obtain individual object instances by using a spectral clustering method. Once the point cloud is partitioned into various point segments corresponding to object instances such as single trees, poles or traffic signs, different fine classes can be further distinguished by using local shape descriptor of point segments. After the refinement of the segmentation, vegetation can be separated from other disturbing objects and detected as individual trees. An experimental study shows and evaluates the feasibility and capability of the proposed strategy towards detecting 3D individual trees from MLS data of complex urban corridors.