Stable Binocular Vision Tracking Based on Kalman Filtering with Motion State Estimation
Yinghao Zhang, Zhen Sun, Chunheng Lu, Junchen Wang · 2021
This paper aims to build a high-precision, fast and stable binocular vision tracking system. The proposed binocular vision tracking algorithm can localize a visual marker consisting of at least three X-shaped corners in real time with six degrees of freedom (DoFs). A fast triangle screening algorithm is proposed to improve the calculation efficiency of the template matching process by 61.52%. A Kalman filtering method based on motion state estimation is also proposed to stabilize the 3D tracking of visual markers, which can significantly reduce the fluctuation of tracking and realize smooth tracking with accurate localization. Finally, the localization accuracy of the binocular vision system was evaluated using a commercial laser tracker, and the experimental results showed that the localization accuracy could reach 0.16 mm.