Abstract: This paper described an algorithm based on artificial intelligence widthfirstsearch algorithm for surface reconstruction of unorganized points. From the incremental computing idea, it makes full use of the state expanding characteristic of search algorithm. Recurring to octree space division, searching constraint and optimum vertex estimation, the algorithm uses initialized triangle as searching base and orientation edges as searching elements to reconstruct model surface gradually and symmetrically. The proposed algorithm supports parallel computing for visualization and does not depend much on parameters. In addition, holes and gaps can be filled optionally. The experimental results show that this algorithm is effective, robust and works well for models with arbitrary topology.