A robust and efficient Boolean operation algorithm for point sampled models is presented in this paper. First, a surfel with a certain size of radius is reconstructed at each sample point on the models. And all of the surfels are classified into one of the following categories: in, out and intersect with respect to the other solid model. Then the intersection curves are estimated under the control of a given global error through adaptively subdividing and re-sampling of the intersect surfels. Besides, a hierarchical structure k-d tree is built for each point model to accelerate the test of efficient classifying the surfel’s in/out/intersect test. The experimental results show that this Boolean operation algorithm can robustly handle point models with different sampling resolution and non-uniform sampled point models.