Detection and tracking of targets against a cloudscene background using space-based sensors

This paper describes work that has been done to assess the performance of a three-dimensional velocity filtering algorithm for detecting and tracking a moving target against a cluttered background. The algorithm is considerably more sensitive than conventional frame-differencing techniques, and can detect fainter targets, at the cost of some additional computational complexity. Test images have been generated which contain a target in a cloudscene background, and incorporate typical sensor effects such as noise, blurring (due to both jitter and optical point spread function) and drift. Results are shown in terms of false-alarm rate as a function of sensor noise for different targets, conditions, and wavebands. Some of the trade-offs that have been explored include the choice of either a narrow absorption band (MWIR, 4.22-4.45Cl) for reduced clutter, or a wider band (ML WIR, 5.4±7.2 μ) for increased signal. The optimum band is shown to depend on the sensor noise characteristics and the platform stability