Filtering of Airborne Laser Scanner Data Based on Segmented Point Clouds

The extraction of points on the bare Earth from point clouds acquired by airborne laser scanning is the most time consuming and expensive part in the production of digital elevation models with laser scanning. Current algorithms for filtering point clouds assume the Earth’s surface to be continuous in all directions. This assumption leads to smoothed terrain representations in case of height discontinuities as they are often found in urban environments. This paper presents a new approach to filtering point clouds in which the point cloud is segmented into smooth segments that may still contain height discontinuities. The resulting segments are subsequently classified bare earth or object surfaces based on the geometric relationships with the surrounding segments. The paper demonstrates the advantages of segment-based classification with an analysis of data sets used in the ISPRS filter test.