Abstract :
A modern scanning radar is designed with sufficient sensitivity to detect small targets, and as a consequence will produce plot returns from many clutter objects. Much of the clutter, such as land clutter returns from large fixed objects, will produce a plot on almost every scan and algorithms may be designed to track and remove such clutter. However, many clutter sources such as sea or rain clutter produce plots having little or no correlation in position from scan to scan. If a track is seduced by such clutter plots, or if a false track starts within such an area of clutter, then the track may be maintained using random plots for some considerable time before failing to associate for sufficient scans to cause deletion. Indeed, the instability of the track will reduce the level of track smoothing, increase prediction error and open up association gates to make track maintenance more likely with each update. It is required to identify tracks that are maintained by random clutter; this paper shows that a measure of fractal dimension provides a suitable mechanism for identification.
