2016.P.1.5. Detection and Tracking of Moving Target using Track Before Detect (TBD) method
Naima Amrouche (1,2)
Ali Khenchaf (2)
Daoud Berkani (1)
- National Polytechnic School of Algiers, Algeria
- CNRS, France
Target Tracking, Particle Filter, Track Before Detect, Bayesian Estimation, Monte Carlo Methods, Nonlinear Filtering, Dim targets
The traditional techniques to target tracking are based upon target measurements such as position, range and rate that are obtained through thresholding the outcome of a signal processing unit of a monitoring sensing unit. The main purpose of thresholding is definitely to minimize the data flow and hence facilitate tracking. In order to get a target of a specific signal-to-noise ratio (SNR), the selection of the detection threshold identifies the likelihood of target detection as well as the density of false alarms.
The unwanted influence concerning the thresholding of the sensing unit data lead to restrain the data flow afterwards gets rid of possibly beneficial information.
For high SNR targets this decline of related information is normally of small issue because the desired probability of detection with a little false alarm rate can be achieved. The latest developments in the design of stealth aircraft and military cruise missiles highlighted the issues which needed to detect and track low SNR targets. For these particular dim targets, there is a significant conveniences being used the thresholder data for synchronised detection and track founding. The principle of synchronised detection and tracking applying a thresholder data is recognized in the scientific literatures as track-before-detect (TBD) technique.
Thus, in this paper, TBD algorithm has been proposed to detect and track small moving targets in a sequence radar images. The general proposed methodology is included two steps: firstly, we introduced the target and sensor models. Secondly, we propose the use of TBD approach as a nonlinear filtering problem to describe the conceptual recursive solution. Consequently, the implementation of this solution is based on the use of Particle filter (PF) for different model target motion (Constant velocity, coordinate Turn model). Finally, various simulations were performed and the obtained experiment results show that the Particle filter based on track before detect algorithm (PF-TBD) which is capable to detect a moving target using different models with a low SNR and also, it is able to track the target with a small root mean square (RMS) for position.
- Download the poster in PDF format here (1MB)