Multiple Target Tracking: Fast Algorithm for Data Association and State Estimation.
Abstract
A unified framework is proposed for providing a systematic scheme for generating the data association hypotheses efficiently in the target- oriented, measurement-oriented, and track-oriented approaches to multitarget tracking. A fast recursive algorithm for computing the a posteriori probabilities, suitable for implementation in a distributed multiprocessor system is developed and its links to the theory of permanents is established. An analysis of this algorithm reveals its superiority over existing ones in the average case. In the related problem of direction-of-arrival estimation, a new non-search-type subspace method, called the PESS method, is proposed. This method exploits the structure of the steering matrix more thoroughly to yield a residual-error theoretically shown to be either less than or equal to that obtained by LS-ESPRIT. Furthermore, simulation conducted on several sets of data showed that the PESS method outperforms the TLS-ESPRIT method. Constraints for forcing all roots of a polynomial to the unit circle are obtained for more reliable estimation especially in the low SNR case. Finally, for improved pre-processing to facilitate tracking, a theoretical analysis is proposed to evaluate the robustness of a TLS algorithm, developed earlier, for image reconstruction from a sequence of undersampled noisy and blurred frames.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 21, 1995
- Accession Number
- ADA300870
Entities
People
- N. K. Bose
Organizations
- Pennsylvania State University