Tracking of moving objects is often accompanied with colored process noise (CPN) in the object speed. To reduce tracking errors, the Kalman and unbiased finite impulse response filtering algorithms are modified in thi...
详细信息
Visual target tracking is often accompanied with colored measurement noise (CMN) caused by the object and camera frame dynamics, which require special estimation algorithms. In this paper, we consider the Kalman filte...
详细信息
Tracking of moving objects is often accompanied with colored process noise (CPN) in the object speed. To reduce tracking errors, the Kalman and unbiased finite impulse response filtering algorithms are modified in thi...
详细信息
Tracking of moving objects is often accompanied with colored process noise (CPN) in the object speed. To reduce tracking errors, the Kalman and unbiased finite impulse response filtering algorithms are modified in this paper assuming the Gauss-Markov noise nature. The state differencing approach employed to derive the algorithms, requires solving a nonsymmetric algebraic Riccati equation that allows modifying the system matrix for CPN. Based on a simulated tracking example, a higher accuracy of the modified KF and unbiased finite impulse response (UFIR) filter is demonstrated. Extensive investigations of the walking human trajectories are provided for colored speed noise.
Visual target tracking is often accompanied with colored measurement noise (CMN) caused by the object and camera frame dynamics, which require special estimation algorithms. In this paper, we consider the Kalman filte...
详细信息
Visual target tracking is often accompanied with colored measurement noise (CMN) caused by the object and camera frame dynamics, which require special estimation algorithms. In this paper, we consider the Kalman filter (KF) and unbiased finite impulse response (UFIR) filter modified for Gauss-Markov CMN. The algorithms are designed employing measurement differencing and demonstrate a better performance under the CMN then the standard algorithm. Experimental investigations are conducted for several benchmark trajectories, where a target is tracked using a video camera.
暂无评论