Wavelet-based extraction of building features from airborne laser scanner data

A new approach based on wavelet analysis to detect buildings in a dense urban area from airborne laser scanner data is presented in this paper. Without the spectral reflectance from the buildings, their detection from the laser cloud points is mainly based on the discrimination of the building's elevation and its surroundings. This detection becomes more challenging in a dense urban area, which contains skyscrapers, interspersed with a myriad of low and small as well as large houses along with crowded outdoor human activities. Integration of the object's size and its elevation could mitigate the difficulty of detection. Wavelet analysis was proposed and adopted to build up the framework for size-based detection. The study focused on the detection of buildings and the generation of a three-dimensional (3D) building database. The extractable information from the aerial photographs is optional. The proposed approach was tested in Shinjuku-ku, Tokyo, Japan, and the result has successfully matched with the existing two-dimensional (2D) vector data. Wavelet-based multi-resolution has proved an appropriate approach in eliminating the unnecessary features surrounding buildings and in extracting the buildings.