Proposing A Method Based on Support Vector Machines and Mathematical Morphology to Detecting the Urban Road from Airborne Laser Scaning Data

Today, Airborne laser scan has an important role in acquiring 3-d information from earth surfaces. Manually feature extraction from laser scan data is too time-consuming and expensive. Roads are the most fundamental linear features and extracting of their information is very important for related organizations in each country. The main aim of this research is to present an approach for detecting roads from LiDAR data, which exclusively uses data produced by LiDAR system (Range and Intensity). For this purpose, First the intensity data and then, both of range and intensity data were classified by Support Vector Machine. Next, high objects were removed by using slope-based filtering algorithm and Digital Terrain Model and Digital Non-Terrain Model layers were obtained and Then the result of classification step was modified by making use of Digital Non-Terrain Model layer. Next, by post processing includes five steps of "morphological cleaning", "eliminating small components", "connecting discontinuity in the roads Network", "deleting disconnected components" and "morphological closing" roads were identified from LiDAR data.By comparing the results of implementation the algotithm with ground truth data,values of 84.35% for completeness, 71.61% for correctness and 63.22% for quality were achieved.