Landmark design in 3D PDM for PCA of shoe last

In this paper, the manual and automatic extraction of shoe last landmarks, including landmarks design, are discussed on the base of key sizes which are used in the shoe last design. The effect is evaluated from the following three aspects: the approximation rate of area, the cumulative variance of various landmarks, and the variety of principal shape vectors. Experimental results indicate that the manual approach is more suitable because fewer landmarks are required than the automatic approach. There are three contributions of this paper: 1) critical shape points are provided for shoe last and foot shapes analysis; 2) a method for landmarks selecting and extracting is suggested; and 3) three benchmarks to evaluate the effect are proposed.