In recent years, the particle filter has gained prominence in the area of target tracking because it is robust to non-linear target motion and non-Gaussian additive noise. Traditional track filters, such as the Kalman filter, have been well studied for linear tracking applications, but perform poorly for non-linear applications. The particle filter has been shown to perform well in non-linear applications. The particle filter method is computationally intensive and advances in processor speed and computational power have allowed this method to be implemented in real-time tracking applications. This thesis explores the use of particle filters to detect and track stealthy targets in noisy imagery. Simulated point targets are applied to noisy image data to create an image sequence. A particle filter method known as Track-Before-Detect is developed and used to provide detection and position tracking estimates of a single target as it moves in the image sequence. This method is then extended to track multiple moving targets. The method is analyzed to determine its performance for targets of varying signal-to-noise ratio and for varying particle set sizes.
The simulation results show that the Track-Before-Detect method offers a reliable solution for tracking stealthy targets in noisy imagery. The analysis shows that the proper selection of particle set size and algorithm improvements will yield a filter that can track targets in low signal-to-noise environments. The multi-target simulation results show that the method can be extended successfully to multi-target tracking applications.
This thesis is a continuation of automatic target recognition and target tracking research at Cal Poly under Dr. John Saghri and is sponsored by Raytheon Space and Airborne Systems.
| Identifer | oai:union.ndltd.org:CALPOLY/oai:digitalcommons.calpoly.edu:theses-1230 |
| Date | 01 December 2009 |
| Creators | Losie, Philip M |
| Publisher | DigitalCommons@CalPoly |
| Source Sets | California Polytechnic State University |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Master's Theses |
Page generated in 0.002 seconds