Summary: | 碩士 === 國立中興大學 === 電機工程學系所 === 100 === This thesis proposes a real-time three-dimensional (3D) human body significant point estimation method by using two-dimensional (2D) extracted body contours. The located 3D body significant points include the head, center of the body, tips of the feet and the hands, the shoulders, the elbows, and the knees. This thesis uses one stereo vision camera to capture both the left and right sets of image sequences at the same time. An Angle-Compensated RGB (AC-RGB) segmentation method is used to segment a moving object from the background in RGB color space and reduce the influence of shadow. In each of the left and right images, 2D locations of the head and tips of the feet and the hands are obtained based on 2D contour convex points, body geometrical characteristics, skin color, and temporal information. Locations of the elbows and the knees are obtained by using the skeleton pixels of a 2D body silhouette. The depth of each significant point is obtained by using the 2D locations of the same point found in the left and right images. The Kalman filter algorithm is used to correct locations of the 3D significant body points and predict the points with mismatching locations from the left and right images. The 3D locations of the significant points and the estimated orientation of the human body are sent to the Virtools software to reconstruct a virtual 3D human model. Experimental results show that the AC-RGB human body segmentation outperforms several segmentation methods used for comparison. This thesis sets up a real-time virtual 3D human model system to verify the effectiveness of the proposed approach. To show the potential of the system, the system is also applied to some 3D interactive entertainment games.
|