Mapping Road Based on Multiple Features and B-GVF Snake

Abstract

As a significant application in aerial image, road mapping is still a difficult task since roads show complex features caused by the influence of spectral reflectance, shadows and occlusions. To achieve a satisfying result, a new method combing multiple road features and biased gradient vector flow (B-GVF) snake is studied in this paper. First, an exponential function is applied to fuse the color-based and structure-based measure for gaining the saliency maps which is viewed as the candidate region of B-GVF snake; Secondly, the initial road boundary is calculated from the candidate region using a region-growing algorithm, and then an gradient map is produced by an Gaussian filtering function; at last, a normally biased GVF external force is proposed for mapping road edges, which keeps the diffusion along the tangential direction of the isophotes and biases along the normal direction. Experimental results show that the proposed approach has the good performance in Completeness, Correctness, and F-measure comparing with other state-of-the-art methods.