Vision based in-motion detection of dynamic obstacles for autonomous robot navigation

Autonomous robot navigation in unstructured and dynamic environments requires detection and estimation of the trajectories of moving obstacles so as to perform collision free navigation. This problem is further compounded by the need to detect and track moving objects while the robot is in motion. This poses the challenge of detecting and tracking moving obstacles from a moving observer as against the detection and tracking of moving objects from a stationary observer. In this paper, we propose a method for in-motion detection in dynamic environment. This method is computationally simpler in comparison with the motion recovery achieved from the optical flow based methods. Additionally, the proposed technique has better time response and overcomes the computational overload associated with the use of geometric modelling techniques. It is also robust to low level intensity variations and shadow effects.