Human body tree structure model application in sports techniques three-dimensional reconstitution

Human motion recognition is one of research hotspot in recent years computer vision field, the technique promotes sports techniques development to considerable big extent. Sports researchers tend to regards video sequence images as important reference information, but sequence image is two-dimensional image after being reduced dimensions by video camera, two-dimensional image restricts necessary movement analysis to great extent, so people have urgent expectation in accurate two-dimension images three-dimensional reconstitution. The paper proposes a human model-based multiple views hierarchical image block texture expressed algorithm, in the hope of exploring the algorithm application in human movement three-dimensional reconstitution. The paper focuses on analyzing discriminant model method and generative model method. It provides human body tree structure model and image background elimination process particle filter principle that is required to use. For MH-L1 trackers principle and multi views hierarchical image block texture sparse expressed principles, it makes analysis, explores multiple views hierarchical image block texture sparse expression’s algorithm steps in three-dimensional reconstitution process. On the basis of designing self-sheltering and human model inaccurate calculation caused wrong texture handling algorithm, it applies Matlab software to make three-dimensional reconstitution on four camera images, and displays reconstitution effects.