Abstract: Traditional Procrustes normalization needs many iterations which will spend a lot of time. Here training samples alignment was set only after once translation, rotation and scaling operations by marking anchor point and using average body shape as the initialization rules model. Traditional ASM algorithm leads to a long computing time and is easily to make the feature points matching error for gray model’s similarity. It was improved by using every feature points as a center point, training gray model though its rounded rectangular gray distribution, and searching target points within its 24 neighborhood points. Experimental results show that the key feature point positioning method for Human body based on this improved ASM reduces the number of iterations, shortens the running time, and improves the positioning accuracy.

Key words: human body point distribution model (PDM), active shape model (ASM), improvement of ASM algorithm, feature point positioning