Human Motion Detection Of Two-Dimensional Key Points And Reconstruction On Three-Dimensional Modeling

Along with the improvement of Internet bandwidth and advancement of computer information processing, since multimedia information includes massive data makes human rely on computer to process available digital information. Video itself including lots of available information can be processed by computer is possible, in which the human motion vision analysis is becoming prominent. Recent year, human motion vision analysis is taking leading in computer vision area, and relative application areas include virtual reality, video monitor, perceive interface and so on.Human motion vision analysis is the one of most active research topics in computer vision area, and it aims at human motion image sequence to analyze and process, which usually involves three parts. The first part is human motion detection. It means that human motion area is segmented from static background area in this text. The second part is human motion tracking, including based on model, area and key point. The third part is human motion action and understanding. Among them, the first and second parts are low-level processed and the last is high-level processed in vision analysis area.The article realizes human motion detection, key point detection and human motion three-dimensional reconstruction on video sequences. It includes three procedures in process of realization. The first procedure is human motion segmentation. The way adopted is that using frame-difference to construct a background model to be initialized and updated. As a result of shadow produced by light effects on human motion detection, the paper changes RGB color space information into HSVChue, saturation, value) color space information of hue and saturation to remove the effects of shadows on human motion segmentation so as to human motion detection exact. The second procedure is key point detection. The core technique utilized is APAR(anti-parallel line) extract area idea to detect human motion key point and the labeling. The third procedure is based on Pin-hole principle to transform the two-dimensional image into three-dimensional space of camera labeling mapping, to carry out human motion key point three-dimensional reconstruction at last.