This technology is a high precision method for simultaneously tracking and detecting moving objects. The technology combines a generic object detector that obtains noisy measurements of an object's position with a Kalman filter-based algorithm that enables prediction of both the position and velocity of a moving object in any unconstrained video streams, while minimizing the drift effect.
Object trackers with appearance-based tracking algorithms often exhibit drift effects, in which the estimation precision of the tracked object's position decreases over time. This effect can be particularly conspicuous when tracking objects in unconstrained video sequences with complex backgrounds and transient occlusions. The technology addresses such issues by implementing a Kalman filter-based framework using a generic object tracker and detector. The technology's Kalman filter-based approach is advantageous in that it mitigates the drift effect by simultaneously predicting both the current position and velocity of the tracked object using the noisy measurements of its position.
The performance of the proposed tracking algorithm has been successfully demonstrated using standard video sequences.
Patent Pending (WO/2012/138828)
Tech Ventures Reference: IR M11-085