A point cloud simplification based on the information entropy of normal vector angle is proposed, in view of the difficulty to ensure the optimal of precision and speed of simplification. The principal component analysis is used to estimate the normal of each point and the angle between normal vector and reference plane is computed. The K-nearest neighbor search algorithm is used to determine K-nearest neighbor points, and the local entropy of normal vector angle is proposed according to information entropy. The local entropy represents the features of surface. The point cloud is gradually simplified according to the different local entropy, the more points of convex region are retained and more points of plane are simplified, the non-uniform simplification is realized. The experimental results show that the proposed method can achieve a balance of precision and speed of simplification.