Research On Modeling Of Virtual Human Motion Based On Depth Information

Virtual human movement technology has been widely used in sports andmedical rehabilitation. Performers and real action need to make computergenerated virtual characters, in the film industry or computer games traditionalsystems often requires performers wear with markers, using motion captureequipment to capture and high-quality processing of shooting.Conventional motion capture method requires fixed some of the reflected ortransmitted lightwave functional markers on the point position of the variousjoints of the human body, by simultaneously using multiple cameras to the targetscene sync shooting accurately. In order to improve the capture of motioncapturethe scene, there are some special requirements, such as a plain bluebackground.The paper uses the unmarked point motion tracking method usingbackground subtraction technique depth template extracted from the depth imagesequence, and then the depth template is used to extract the body edge combinedwith synchronized RGB image. By using depth template can more accuratelyextract the information of the edge of the body part. Splitting process does notneed to strictly limit capturing environment, such as a single color background orclothes. Moving target pose in front of the camera calibration can automaticallycalibrate the human part of the region. In this paper, an improved Cam-shiftalgorithm to the color of the target surface distribution characteristics of thehuman body’s major joints point tracking. Assuming that the center of the track toeach of the joint point of the rectangular area for the joint location, using aparameter limit of the proportion of the orthogonal projection of the scale factor,to obtain the body of the joint point of the three-dimensional coordinatescorresponding to each frame. Re-use of the limb joints point acral track dynamicalgorithm, make these joints sharpest motion is smooth. Finally, we use the the parametric inverse kinematics algorithm so that the the virtual movementexperimenter real action is closer to the. Complete virtual human motion model.Finally, this paper uses the proposed method in Visual Studio2010development platform. Accurately tracking of the major joints of the human bodyat a rate of30frames per second, and the establishment of a virtual humanmotion model. The experimental results on the accuracy and speed of analysisand the results show that the model can meet their daily lives on the speed andaccuracy of human motion tracking requirements entertainment.