The paper integrated genetic algorithm and marching method into a novel algorithm to solve the surface intersection problem. By combining genetic algorithm with local searching method the efficiency of evolution is greatly improved. By fully utilizing the global searching ability and instinct attribute for parallel computation of genetic algorithm and the local rapid convergence of marching method, the algorithm can compute the intersection robustly and generate correct topology of intersection curves. The details of the new algorithm are discussed here. The algorithm have been implemented in a prototype system named TigerSurf based on Windows/NT platform, and a soundly result is gotten from test datum.