CFAR detection in heterogeneous K- distributed sea-clutter background

Detection of targets at sea is challenging due to unwanted echo returns from the sea surface, i.e. sea clutter returns. To account for the undesired effects of sea clutter at the receiver, and to control the probability of detection and false alarm, the K-distribution has often been used to provide a promising statistical fit to real clutter data. However, controlling the performance of the receiver becomes very challenging in heterogeneous clutter, that is when there is a sudden transition from one clutter region to another with a change in shape and/or scale distribution parameters. A possible solution to this challenge is to use some prior information on the sea clutter characteristics to generate clutter maps that inform adaptive detection solutions. This prior information can be obtained by the radar in real time (or close to real time) using oceanographic models, statistical clustering or potentially Artificial Intelligence.This research presents our first step in this direction by investigating detection in heterogeneous fully correlated K-distributed sea clutter. A transition line between homogeneous clutter regions is estimated using the statistical parameters of a K-distribution, to avoid polluting the training windows of a Constant False Alarm Rate (CFAR) detector with non-representative data. The transition cells assist to resolve the heterogeneous clutter into small homogeneous clutter regions and for every homogeneous region a CFAR detector is designed according to the K-distribution shape parameter. Results are obtained and presented for simulated data as well as for real sea clutter data provided by Hensoldt UK.