Road Extraction Method at Pixel and Object Level using High Resolution Images and LIDAR Data with Evidence Theory
Rodriguez-Cuenca, Borja; Martinez de Agirre, Alex; Alonso, Concepcion; del Val, Alberto
Alcalá University, SPAIN
LIght Detection and Ranging (LIDAR) data provides accurate height information for objects in the earth surface. It is common to combine LIDAR data with satellite or aerial imagery to determine the location of different cartographic entities. In this work an automated method for road extraction in urban areas from high resolution aerial images is presented. This method is based in Dempster Shafer evidence theory, it consists in fusing information from different information sources. The proposed method is applied to the road extraction in two different ways: at pixel level and at object level. Results provided for both levels are compared with a ground truth created by the authors of this work, to thereby determine which method provides the best accuracy. As input data an aerial/satellite image and LIDAR data are required. RGB and NIR bands of an aerial or satellite high resolution images are used to compute different decision indexes. LIDAR data are used to generate a digital terrain model (MDT), a digital surface model (MDS) and a lidar intensity band, which is used to determine the location of the roads.