Edge Tracing Based on Improved Genetic Algorithm

Abstract

In this paper, we proposed a new edge tracing method with high robustness to noise. Through representing edge with maximal gradient path encoded by chain code, the edge tracing problems can be converted into combinatorial optimization problems, and so they can be solved by genetic algorithm. We optimized the traditional genetic algorithm in order to improve the convergence rate. Our method is effective to edges with any shape because it does not require any prior knowledge about the edges. In this paper we also discussed the problem of edge winding and folding and expatiated how to avoid it by designing proper gene coding method and punishment function. Furthermore, by transforming the region of interests from Cartesian coordinates to polar coordinates before edge tracing, this method can be used for closed edges. The experimental results show this is an effective edge tracing method with high robustness and flexibility.