Vision based obstacle detection and collision avoidance is a major challenge in unmanned vehicle navigation. The gradient based Horn and Schunck algorithm has been optimized for computation of Optical Flow (OF) vectors for both virtual and real images. Based on extensive and systematic simulation studies, the desirable values of important parameters such as Compensation parameter, number of iterations and resolution of images that determine the accuracy and efficacy of Horn and Schunck Algorithms have been suggested.
A novel and computationally efficient Cluster based method is proposed for the determination of FOE. The proposed Cluster based method overcomes the earlier drawback of rapid variations in the co ordinates of FOE when successive frames are considered for simulation. Separate decision logics appropriate for UGV and MAV navigation have been identified.
The simulation studies reported in this paper generally treat both the synthetic as well as real images.
Micro Air Vehicle [MAV], Unmanned Ground Vehicle [UGV], Optical Flow [OF], Focus Of Expansion (FOE), Time To Contact (TTC)
