Dynamic Node Distribution in Adaptive Snakes for Road Extraction

In this paper we address the issue of road extraction from digital imagery using deformable contour models (snakes). We present a novel variation of the traditional snakes solution, where additional nodes are inserted, and redundant nodes are deleted, to better describe the complexity of the extracted line. Node insertion and deletion are based on an analysis of the energy terms of the snake solution. This allows us to use more, closely spaced nodes along the high curvature areas of a road, compared to the linear segments of the same road outline. This dynamic manipulation of the number and spacing of nodes within a single snake allows us to better capture the geometry of the road, and to better accommodate its radiometric behavior. Thus it produces more accurate results than the traditional snake solution. Here we present our approach and experimental results to demonstrate its performance for road extraction in geospatial applications.