Automatic Extraction of Buildings and Roads using Combination of Airborne High resolution Laser Scanner and Multispectral Data

要旨

High-resolution airborne Laser scanner (Light Detection And Ranging) data in conjunction with high resolution airborne multispectral optical data have been used to extract buildings and roads in urban area. The geometric characteristics like shape of buildings and roads have been used in the feature extraction process. The height of buildings and length of the roads have been estimated using Laser scanner data. The paper discusses about the automatic extraction of shape, height and length using geometric properties of buildings and roads for their recognition and reconstruction. The new approach followed here was the use of Laser scanner and multispectral data in combination rather than independently. Data filtering techniques were applied to segment the buildings. For extraction of roads, Gaussian pyramid and edge detection algorithm were applied to extract centre line and roadside parallels of roads. The accuracy of building extraction was about 94%. The approach presented here has advantage over manual interpretation of buildings as because 220 buildings out of 231 were detected through automatic approach whereas through visual interpretation only 201 could be detected. Road length estimated through the analysis is 10.166 km. The study indicates high potential for extraction of complex buildings and road network in an urban environment with high accuracy in cost effective manner.