Research And Implementation Of Non-contacting Anthropometric Algorithm Based On Image

The non-contacting anthropometric algorithm which is based on image,has the advantages of celerity,convenience and low cost.It has become the research tendency of anthropometric technology.The precision of anthropometric measurement depends on the accurate extraction of feature points and the circumference fitting method,and the extraction of feature point depends on the extraction of body contour.This dissertation summarized existing contour extraction algorithms and feature point extraction algorithms,then contrasted and analyzed the merits and demerits of these algorithms through simulation experiments.For the anthropometric scenario,this dissertation pointed out that the existing anthropometric algorithms were strict in the measuring environment when extracting the contour and lacked body information in the process of feature points extraction which resulted in low accuracy.Finally the improved algorithm was proposed.The main contents of this dissertation were as follows:(1)SE-U-Snake Model: Aiming at dealing with the problems existing in the traditional Snake model,SE-U-Snake Model was proposed.It solved the problems of being sensitive to initial contour,being vulnerable to other strong edges,and being insensitive to local contour extraction by means of automatically initial contour setting,shape estimation and the unclosed model.Experiment demonstrated that this model was more suitable for anthropometric local contour extraction.(2)Anthropometric algorithm based on body proportion and local contour: In order to improve the defects of existing anthropometric algorithm,this chapter integrated prior knowledge,such as the body proportion,into the algorithm.Then divided human body image into characteristics areas,where the local contour and feature point were extracted.Finally calculated the shoulder length and the arm length by measurement scale,and calculated the neck circumference and the chest circumference by binary linear equation.Experiment demonstrated that the average error of the proposed algorithm was within 5%,which met the anthropometric requirement.(3)Design and implementation of anthropometric system based on image: This dissertation combined object-oriented technology with the hierarchical thought to design the system architecture.Then built OpenCV developing environment and MySQL running environment,implemented the system and completed the actual measurement of length parameters and circumference parameters.Finally the measurement results were showed through a visual interface.