Neural network-based adaptive radar detection scheme for small ice targets in sea clutter

An adaptive radar detection algorithm for the detection of small pieces of ice floating in an open sea environment is presented. The detection is carried out in the time-frequency domain, which eliminates the limitations of standard Doppler processing. Instead of looking for the ´best´ time-frequency representation, the detection scheme is based on the ambiguity function and the focus is more on extracting information needed for classification. The detection scheme uses a multilayer perceptron as a pattern classifier. Using actual radar data, the adaptive detection results are compared with a traditional time-frequency domain constant false-alarm rate (CFAR) processor.