3D object-based classification for vehicle extraction from airborne LiDAR data by combining point shape information with spatial edge

The problem of vehicle extraction using airborne laser scanning (ALS) is studied under the framework of object-based point cloud analysis (OBPA). Object extraction relies on the partitioning of raw ALS data into various segments approximating semantic entities followed by classification. A 3D segmentation method working directly on point cloud is used, which features the detection of local arbitrary modes and the globally optimized organization of segments concurrently. To make the segmentation more competent in extracting small-scale objects such as vehicle, the detection of local structures is realized by adaptive mean shift (MS) using variable bandwidths which are determined by the point shape information bounded by spatial edge. The experimental results show that the proposed method performs very well in terms of visual interpretation as well as extraction accuracy.