A Rule Based Analysis of Image Objects for Road Detection in Urban Areas Using High Resolution Satellite Image and LiDAR Data

Roads detection in urban areas is of greater importance and is constantly considered as a main issue of researches in remote sensing community. According to spectral and geometrical variety of road pixels as well as their spectral similarity with other features such as buildings, parking lots and sidewalks and sometimes discreteness of roads due to having vehicles and trees in their neighborhood makes it problematic to precisely identify urban roads through satellite images. On the other hand, LiDAR data, providing height information, make it possible to recognize roads from other spectrally similar elements. Therefore, it has been used in many researches to identify different features like roads along with satellite images. In this paper Quick Bird large scale satellite image and LiDAR data through an object oriented analyses have been used to extract various types of urban roads. Proposed method has designed and implemented a rule oriented strategy based on a masking strategy. Afterwards, a supplementary strategy based on conceptual features design was used. The overall precision of class identification is 89.2 % and kappa coefficient is 0.832 which show a satisfactory precision according to different conditions and many interclass noises. Final results demonstrate high capability of object oriented methods in simultaneous identification of wide variety of road elements in complex urban areas using both high resolution satellite image and LiDAR data.