Real-time Tracking Of The Behavior Of Three-dimensional Reconstruction Of The Moving Hand

The gesture tracking and recognition of the behavior of the moving hand based oncomputer vision is a very active area of the field of human-computer interaction research, andis also a hot issue in recent years. In the tracking method of the behavior of the moving hand,the main task of the model-based tracking methods is to continuously get the posture andposition of the three-dimensional hand corresponding with per frame image, and this methodis an important idea in hand tracking research areas. There exist problems that the dataprocessing time is too long for each frame and the real-time nature of the interaction is notstrong and so on in the present methods.For the problems in tracking methods such as the real-time nature of the interaction isnot strong, we put forward a new method. By using the interframe consistency principle, weuse the voxel-based visual hull method to rebuild the three-dimensional hand, and get itsskeleton model, and then do the real-time tracking on the moving hand.1. Firstly, we do reseach on the existing modeling ideas and methods based on the voxelvisual hull, and conduct real-time three-dimensional reconstruction by using the video imageobtained by the multiple cameras of the moving hand, and finally get the voxel visual hull ofthe moving hand. Study GPU-based parallel processing algorithm, and optimize the efficiencyof modeling process, to improve the efficiency of the system.2. The paper studies the skeleton extraction algorithm of the voxel visual hull, and basedon the assessment of the advantages and disadvantages of various skeleton extractionalgorithms, chooses an efficient and accurate one for the moving hand. The method makes fulluse of the interframe consistency principle, and conducts real-time and efficient skeletonextraction for three-dimensional model, in order to achieve real-time tracking of themovement of the moving hand.3. We do reseach on the tracking and recognition algorithms of the behavior of themoving hand, and analyse the characteristics of moving hand, and analyse and discriminantthe behavior of the moving hand by using the skeleton model, and determine the graspbehavior of the hand, and complete the collision detection, so as to realize the virtual realityinteraction. We realize the algorithm put forward in this paper on the simulation platform ofreal-time tracking of the movement of manpower behavior, and analyse the data of time withthe existing methods. The experimental results show that, the real-time tracking algorithm putforward in this paper can improve operational efficiency, and enhance the real-time nature ofthe interaction, in the premise of guaranteeing run effect.