Abstract: In order to recognize and extract feature dimensions automatically from point clouds of human body,a new method based on fuzzy rules is proposed.First,point clouds direction is adjusted to X-axis according to the width of the points projected to the knee plane.Second,a plane moved along Z-axis with variable steps to search the human body clouds.At the mean time,feature points positions are recognized by the fuzzy rules based on the changed parabola membership function and coordinate value comparison rules.Finally,critical feature dimensions are calculated by Z-coordinate difference method and intersection contour perimeter methods.Experimental results show that the proposed methods are feasible and reliable in practice.