Improvement in the Performance of Multi-Scan Target Tracking Algorithms using Fixed Lag Smoothing

Abstract

The efficient tracking of targets in the surveillance area in the presence of other objects (clutter) is the key issue in a target tracking problem. Fundamentally, the target state is hybrid in nature and is determined by both the target existence state and the target trajectory state. In the presence of clutter and difficult surveillance conditions, the estimation of the target hybrid state degrades due to the various parameters that affect the performance of the tracking algorithm. The multi-scan data association algorithm proposed in this thesis provides a sliding window based approach for data association, which limits computational resources and improves the target hybrid state estimation in a heavily cluttered environment, where the target is present with a low probability of detection. Due to the on-line nature of the filtering algorithms, their performance has certain limitations. To solve this problem, a proper smoothing procedure can provide an improvement in the tracking performance under difficult surveillance conditions by providing the revised target hybrid state at a past scan using the measurement information till the current scan. The work presented in this thesis proposes a new smoothing algorithm for target tracking algorithms. The smoothing algorithm proposed in this thesis provides a general method to smooth the target hybrid state at any fixed lag of size N in the presence of clutter. The proposed work also provides an augmented state smoothing architecture for various target tracking algorithms, which is a key part of the proposed fixed lag smoothing algorithm. The extensive simulation study carried out in this thesis provides a comprehensive overview on the performance of smoothing target tracking algorithms. Different simulation scenarios are considered to provide a better environment for performance assessment. A detailed analysis regarding the performance of smoothing algorithms is also added , which compared to the on-line tracking algorithms provides a significant improvement in terms of target hybrid state estimation.