Feature Based Plot Classification using a Bayes Algorithm

Many fielded state-of-the-art radars suffer from excessive false alarm rates, at least at times. The radar concerned in this paper is intended for automatic tracking of objects for maritime surface surveillance, one of the difficult scenarios for automatic tracking. The algorithm first corrects for the ship\´s motion and then builds a plot map, which stores the long term probability density functions of features of the observed echoes. Through the Bayes algorithm the conditional probability that a plot belongs to each of the four classes "moving clutter", "fixed clutter", "ship" or "other object" is established. The result is a significant reduction of the clutter plots and only a modest reduction of the ships\´ plots. Although the method presented here is intended as a classifier for prioritizing plots in the case of data overload, it could also be used in the initiation phase of new tracks in order to classify tracks in either of the mentioned classes