Improved Detection and Tracking of Small Targets in a Cluttered Environment

摘要

Robust automatic detection and tracking of small targets in cluttered environments is becoming increasingly important; this is especially true in the surveillance of areas of high strategic importance. This paper describes an unattended electro-optical tracking system, designed to automatically detect and track moving targets in cluttered environments. Such a system has to have a low false alarm rate whilst maintaining a high probability of detection. Once a target has been detected a security alert can be issued and the security personnel are automatically shown the relevant images and a risk factor can be applied to each target. Often surveillance systems will be positioned such that there will be sources of false alarms within view. Rejection of these sources is critical, however rejection of genuine targets close to them should be avoided. Methods of rejecting clutter are investigated, these include rejection of known features e.g. vegetation, and rejection of targets that conform to expected patterns e.g. vehicles on a road. Further to this, target tracks are maintained in a 'local' Cartesian co-ordinate set allowing the possibility of pin-pointing the target and maintaining a track whilst scan capability.