A New Spherical-target Tracking Method Based on Monocular Vision

摘要

Target tracking technique is increasingly employed in Virtual Reality (VR) products and monocular vision tracking technique is the key research fields. According to VR products development requirements, this paper proposes a new spherical-target tracking method which is based on monocular vision. This tracking method based on worm model contour detection algorithm, and modified the worm motion rule in order to decrease the detection period and facilitates the stability of object contour detection. Nelder-Mead optimal algorithm is also included in this tracking method which can improve the location of spherical objects under the circumstances of partly overlap and meet the demands of multimodal interaction requirements. The test results show that this method can conduct effective real-time tracking to multi-targets and possess high abilities of resistance to interference.