On-board camera 3D-motion analysis

Abstract

Estimation of on-board camera's 3D motion has been a very important technique for autonomous robot navigation and other vision-based detecting and measuring tasks. Due to the nonlinearity and continuously changing background, on-board camera's 3D-motion analysis is very complex and always employs optical flows, line matching of road lane-marks or horizon tracking. It incurs a huge calculation cost and is very sensitive to noise. A new estimation method with the introduction of camera's focus of expansion (FOE) is presented in this paper. Using the motion features of the on-board camera and FOE, the camera 3D-motion parameters can theoretically be determined by our method using only three matching pairs, which make it fast and more efficient for real-time applications compared with the traditional methods. Experiments on simulation images and image sequences of real road scenes show the effectiveness and precision of the proposed approach.