A Kalman filtering-based dynamic programming track-before-detect algorithm for turn target

Abstract

Dynamic programming-based track-before-detect algorithm is an effective approach to track target with uniform rectilinear or slowly maneuvering motion. But when target performs a turn motion it has a poor performance for the transition step is fixed. In this paper a new algorithm which combines dynamic programming with Kalman filtering has been proposed. This algorithm makes use of state prediction operation in Kalman filtering to change the transition step in DP algorithm adaptively. Numerical results show that this new method has a much better tracking ability than the traditional procedure when target turns.

Remote Sensing Low Signal-to-Noise-Ratio Target Detection Enhancement

“…We have demonstrated in a modern PC with a GPU card that MMODS can achieve real-time performance with modern megapixel cameras that detect a large quantity of targets. In comparison, most existing techniques can only support the simultaneous detection of a low number of targets [ 14 , 15 , 16 , 17 , 18 , 19 , 20 ]. It is impossible for a human eye to monitor 10 million pixels at a given time, but MMODS can do all this in real time.…”

“…This method assumes a known number of targets and its constant motion trajectory. The Track-Before-Detect (TBD) approach [ 15 ] incorporates the use of dynamic motion modeling, such as the Kalman Filter [ 16 , 17 ] and the Particle filter [ 18 , 19 ], to predict and integrate target motion over multiple frames to enhance a target’s SNR. However, the TBD approach is generally limited to slow moving targets as performance tends to degrade for high-speed targets [ 20 ].…”