Contour Matching Based On Shape Contribution Factor And Its Application In 3D Human Reconstruction

Finding meaningful matching between shapes is a basic problem in Geometry Processing.In the field of shape matching,people are accustomed to regard objects as a set of discrete plane points,and contour is the boundary line to construct any shape.Compared with point sets,it has more order relationship of object boundary,and is a higher level of visual information.In recent years,the contour-based shape matching method has achieved rapid development,among which the shape context matching algorithm is the most classic.However,there are still many problems in this field that have not been solved.For example,most of the work of shape matching focuses on finding the similarity between shapes,and does not pursue the precise matching between contour point sets.It allows the existence of abnormal matching point pairs in the matching results.Therefore,how to quickly and effectively deal with abnormal matching point pairs in matching results is worthy of further study.Aiming at the shortcomings of shape matching,this paper proposes a contour matching method based on shape contribution factor on the basis of in-depth study and summary of shape matching algorithm and traditional shape context algorithm.The proposed shape contribution factor is mainly used to evaluate each matching point pair in the matching result,which describes the influence of a single matching point pair on the whole matching result.Combined with the sequence information of contour and nearest neighbor information,the abnormal matching points in the results are quickly processed.In this paper,the proposed contour matching method is successfully applied to the matching of human shape depth data based on kinect depth camera,and the fast and effective 3D human body modeling in any posture is carried out.The 3D human body model reconstruction system designed in this paper uses a single human body image to estimate the posture and shape of the deformable model.Posture reconstruction mainly uses the human skeleton data obtained from the image.The key of shape reconstruction is to obtain the accurate correspondence between the projection contour of the 3D human model and the human contour in the image.Finally,this paper conducts a comparative experiment with other highly correlated contour matching work on the self-made data set.The experiment proves that the shape contribution factor proposed in this paper can effectively deal with the abnormal matching point pairs in contour matching,so as to obtain a more reasonable contour correspondence visually.