Wavelet feature extractors and evidence neural networks for the detection of “blackholes” in backscatter ionograms

Backscatter ionograms provide a concise summary of high frequency (HF) ionospheric propagation conditions, and find application in over-the-horizon radar (OTHR) and HF communications. Of particular interest are blackholes in backscatter ionograms, which are manifest as anomalous linear depressions in the ionogram clutter. An algorithm is developed for the real time automatic detection of blackholes. The algorithm uses a wavelet based feature extractor derived for a simple blackhole model, which in turn is presented as input to a multistage evidence based neural network classifier. Classification results are favourably compared with classical template matching