Scene Perception and Classified Detection for Roads in Remote Sensing Images

This paper presents an improved perception model for roads in remote sensing images based on the principles of perceptual organization and classification fusion in HVS.The model consists of four levels： pixels,elements,structures and objects,and two additional sub-processes are incorporated compared with the traditional one：Automatic classification of road scenes and global integration of multiform roads.Based on the model,a novel algorithm for detecting roads from remote sensing images is also proposed,in which two types of road primitives,namely blob-primitive and line-primitive are defined,measured,extracted and linked using different methods for dissimilar road scenes.A hierarchical search strategy driven by saliency measurement is adopted in both linking processes.Finally,all the linked road segments are normalized with center-main lines and integrated into global smooth road curves.Experimental results show that the algorithm can detect multiform roads from real satellite images with high adaptability and reliability.