Vehicle trajectory prediction using a catadioptric omnidirectional camera
Vigneshram Krishnamoorthy, Saksham Agarwal, K. Subramanian Venkatesh · 2016
A practical method is presented to predict the future spatial-temporal trajectories of multiple vehicles at road intersections in real time using a catadioptric omnidirectional camera equipped with an Equiangular mirror. Tracking is done using CamShift algorithm running alongside a Kalman Filter to handle occlusions. Domain transformation of the tracked objects location and velocity from image space to real world is done using a geometrical model. A computationally effective model for trajectory prediction has been presented along with the experimental results obtained using it. Applications such as collision prediction and vehicle tracking or any other event of interest using a dual-camera system are also discussed briefly.