Three-dimensional shape automatic partition method based on Mean Shift

The invention provides a three-dimensional shape automatic partition method based on Mean Shift. Local features of each three-dimensional grid are obtained through distance of a characterization peak, a central coordinate of the grid is obtained through the coordinate of the peak of the three-dimensional-shape grid, and a combined four-dimensional feature space is obtained. Mean Shift algorithm is adopted for conducting cluster calculation in the four-dimensional feature space, and a cluster number and feature points contained in each cluster are obtained. K-nearest neighbor classification technology is adopted for conducting decision space modeling on cluster results and revising partition results partially. Visualization technology is adopted for conducting shading on the partially-revised partition results through a method of marking colors according to cluster attributes. A Princeton partition standard is adopted for calculating standard errors of the partition method under different measurements so as to