Research Of Hand Tracking Technology Based On Active Vision With The Parallel Robot

Flight simulator plays a decisive role in the field of civil aviation, however, the price andmaintenance costs are very expensive. Lightweight simulator based on virtual reality is an attractiveand effective low-cost solution. In the process of HCI of civil aviation, how to reconstruct the handmovements in the virtual scene is critical. Computer vision tracking scheme can provide a naturalnessand non-contact way to obtain hand features, but the existing constraint--self-occlusions of the hand,will affect the effect of hand reconstruction. This paper proposed using active vision technology tosolve the self occlusion problem in the process of the hand tracking.This paper summarizes some examples of the solutions using active vision technology, andpoints out the deficiencies of those solutions. According to the inherent characteristic of parallel robot,the article proposes that the parallel robot Stewart can be used for active vision and, puts forward anew active vision hand tracking technology.In the following chapters, the paper designs the actual platform in the visual servo system,including mechanical parts, mechanical and electrical control part and the video acquisition modules.The basic software of the active vision system is realized, including data acquisition, communication,HCI module. An actual running and debugging platform for the active vision system is finished.According to the overview, principle and classification of the visual servo system, a variety ofvisual servo system is simulated and analyzed. The image-based visual servo technology is chosen tofor the robot control algorithm. Meanwhile, Mean-Shift method is used for target tracking, and how touse the bare hands and the color marking for tracking characteristics is illustrated.Above all, general visual controller and the hybrid controller which considers the dynamiccharacteristics of robot for the visual servo system are designed, and stability analysis of the wholesystem is also carried out via Lyapunov function. On this basis, respective simulations andexperiments are carried out. The results and experimental data illustrate that the active vision systembasing on Stewart can effectively solve the problem of hand occlusion in the hand tracking process.Finally, the shortcomings are introduced, and the research direction of follow-up work is pointedout.