Abstract: Traditional algorithm can avoid color difference areas effectively by using seamlines detection in overlapping color difference image.. However, high density of urban buildings using this method is not good. In this paper, we distinguish ground regional and non-ground area accurately by processing high-precision DSM. Most of traditional seamlines detection algorithms are based on iterative calculations, and have high complexity. Our paper propose a new algorithm named greedy snake algorithm. The new algorithm is only associated with three parameters: searching step, direction rotation interval and region width. We carried out experiments for urban areas for orthophoto seamlines detection. The results show that the seamlines can avoid buildings well, and the greedy snake algorithm can be applied in the optimal path detection effectively.