Abstract: Aiming at the feature point detection problem of 3D models, we propose a novel feature point detection algorithm based on the Laplace-Beltrami operator. For a given 3 D mesh model, firstly the discrete Laplace-Beltrami operator is constructed, and its eigenvalues and eigenvectors are calculated. Then the local extremum and saddle points of the eigenvectors are detected and blended using the weighted formula based on the eigenvalues. Experimental models are selected from SHREC2010, and the feature points are visualized using OpenGL on VS2013. The experimental results demonstrate that our method gets ideal detection results, is applicable to models with noise, and obtains highly similar detection results on isometric models and can process large-scale data through the distributed computing.