Watershed-based Filtering for Object Separation from Airborne LIDAR Data

One conventional method of topographic data collection is image data such as satellite images and aerial photographs. However, due to the low spatial resolution and time-consumption of the conventional data collection process, all the methods that use image data have limitations. The recent advances in the research on airborne Light Detection And Ranging (LIDAR) data enable the collection of more accurate and fast topographic measurements. This technique has the potential to overcome several bottlenecks imposed by the conventional methods. LIDAR data is acquired by the system which usually returns a cloud of irregularly spaced 3D points containing horizontal coordinates (x, y) and elevation coordinate (z) from reflective objects scanned by the aircraft-mounted lasers underneath the flight path. In this paper, a powerful framework based on the watersheds algorithm is developed to detect the nonground measurements. The objects segmented by our proposed framework are then classified. Several areas with diverse feature content, including mountains, urban area etc., are selected to test the watersheds filter using visually qualitative comparison and numerically quantitative comparison. The results show that our proposed watersheds filter accurate in the generation of DTMs in landscapes of high complexity.