A New Region Growing-Based Method for Road Network Extraction and Its Application on Different Resolution SAR Images

Abstract

Road network extraction plays an irreplaceable role in the applications of synthetic aperture radar (SAR) images. In this paper, we propose a new method based on the region growing to quickly extract the road network, which is suitable for different resolution SAR images. First, a weighted ratio line detector (W-RLD) is proposed to extract road features. Then, an automatic road seeds extraction method, which merges the ratio and direction information, is utilized to improve the quality of the extracted road seeds. Finally, the region growing concept is adopted to construct the road network, and a fast parameter selection procedure is presented for adaptively adjusting growing parameters. In experiments, four kinds of SAR images are used to assess the performance of the proposed method, including Envisat ASAR (30 m), HJ-1-C (5 m), TerraSAR-X (3 m), and airborne C-band data (0.5 m). Both visual and quantitative evaluation results show the adaptability and efficiency of the proposed approach.