3D Modeling Method Study Of Human Thoracic And Abdominal Surface Using Binocular Vision

3D reconstruction of human thoracic and abdominal surface which is based on binocular vision has the advantage of simple system structure,strong real-time characteristic,low cost and high efficiency.The 3D measurement of the respiratory motion on the thoracic and abdominal surface of the human body can be applied during radiotherapy,which plays an important role in the analysis and prediction of human respiratory motion.Therefore,this paper has important research values for the 3D of human reconstruction of human thoracic and abdominal surface which is based on binocular vision.This paper mainly focuses on the aspects of parameter design of binocular vision,feature projection pattern design and stereo matching.First of all,in order to improve the measurement accuracy of the binocular stereo vision system,the structural parameter model of the system is established.Then we analyze the relationship between the structural parameters of the binocular stereo vision and the measured object.We can get the range of the parameters where the system error is minimal through simulation experiment.Secondly,in order to solve the problem of few features on the thoracic and abdominal surface,the method of artificial projection of feature points is adopted to increase the number of feature points on the thoracic and abdominal surface.In order to ensure the global uniqueness of the encoded pattern,we use the gray value of R,G and B channel of the color encoding points as three features to design the projection pattern.The experimental results show the feasibility of the projected features designed in this paper.Finally,for the 3D reconstruction of static objects,the improved SURF algorithm is used to improve the accuracy of stereo matching.the six-dimensional RGB color components containing the color information are added to the established descriptors,and int can improve the accuracy of feature matching.In order to solve the stereo matching speed problem in the dynamic 3D reconstruction of the thoracic and abdominal surface,a Kd-tree and BBF search-matching algorithm is proposed to construct a KD-tree structure for the extended feature description vector,and then perform the BBF search to achieve fast matching.The experimental results show that the mathcing time of this paper's method is reduced by 2/3 compared with the method which is based on KD-tree.We can achieve the dynamic stereo matching of thoracic and abdominal surface fast and accurately using the improved algorithm,and get the 3D information of the matching points.According to the principle of parallax measurement,the depth image of the thoracic and abdominal surface is obtained.