Abstract:

Ridge lines can be categorized into main veins and branch veins. Main veins are continuous, prominent and long, which dominate  the overall characteristics and trend of the terrain. Branch veins are relatively shorter and they locate on both sides of the main veins, which reflect the details and variety of the terrain. The results of the existing ridge line extraction methods not only contain too many branches but also suffer large deviation and poor continuity, so these methods cannot be applied to practical applications. We therefore propose a new method for extracting main veins using multiscale analysis and the fusion theory. We firstly use the Gaussian kernel function to decompose digital elevation model data into different scales. Then the ridge lines of all scales are extracted and their weight values are constructed with Gaussian scale parameters. Finally the main veins of each scale are fused by the weighted average method. Experimental results show that the proposed method is simple and feasible, and the main ridge lines extracted by the proposed method are clear and  continuous, and they conform to the topographical features of the terrain.

Key words: digital elevation model;scale space;ridge lines;fusion;main ridge lines