An improved point cloud simplification and reconstruction method is proposed to solve the problems of feature disappearance and hole phenomenon in point cloud scanning and sampling of thin-walled blades. Firstly, the outer contour of the blade is extracted according to the angle threshold of the normal vector and the Euclidean distance. Secondly, the average curvature and Gaussian curvature of the point cloud are calculated, the threshold is set to divide the point cloud into molecular sets, and the index space method is used to simplify the point cloud data. Then, the greedy algorithm is used to analyze the blade point cloud for the distance threshold coefficient and triangle parameters, and the empirical value of greedy triangle parameters without holes is obtained. The relationship between the distance threshold coefficient and the average distance of point cloud is fitted to realize the automatic adjustment of greedy reconstruction parameters. Compared with the other two methods, experimental results show that, when the overall reduction rate is about 90% , the standard deviation is reduced by 26. 45% and 19. 92% , and the average deviation of the outer contour dimension is reduced by 79. 81% and 47. 97% , respectively. The reconstruction parameters are set according to the proposed method, and the reconstruction quality is good, which has a good application reference for the realization of intelligent machining detection of thin-walled blades.