Automatic Vehicle Extraction from Airborne LiDAR Data Using an Object-Based Point Cloud Analysis Method

Automatic vehicle extraction from an airborne laser scanning (ALS) point cloud is very useful for many applications, such as digital elevation model generation and 3D building reconstruction. In this article, an object-based point cloud analysis (OBPCA) method is proposed for vehicle extraction from an ALS point cloud. First, a segmentation-based progressive TIN (triangular irregular network) densification is employed to detect the ground points, and the potential vehicle points are detected based on the normalized heights of the non-ground points. Second, 3D connected component analysis is performed to group the potential vehicle points into segments. At last, vehicle segments are detected based on three features, including area, rectangularity and elongatedness. Experiments suggest that the proposed method is capable of achieving higher accuracy than the exiting mean-shift-based method for vehicle extraction from an ALS point cloud. Moreover, the larger the point density is, the higher the achieved accuracy is.