Research On The Secondary Generation Of Virtual Human Body Based On Customer Shape In 3d Fitting System

Through the single parameter control, the virtual person model shape change rule in V-Stitcher is found. According the rule of shape change, put shoulder slope, chest Position, chest peak distance, abdomen, body width, hip width, upper body length, posture, para posture, Bottom Position, upper body posture parameters which from the 3 d fitting software as the output vector, the human body characteristic parameters as input vector, establishing the multiple input and single output four layers BP neural network respectively in matlab7.1.On the basis of the training results, setting the transfer function for {tansig, logsig, purelin}, training function for {trainlm}, training step length for 1000, Ir for 0.01. Using MIV to chose the most effective parameters, to establish the multiple input and single output network structure. Compeared with the model from 3d scanner, the result indicates that this method can simulate shape feature of the human body well.