ADAPTIVE WAVELET CLASSIFICATION OF ACOUSTIC BACKSCATTER AND IMAGERY

The utility and robustness of wavelet features is demonstrated through three practical case studies of detecting objects in multispectral el ectro-optical imagery, sidescan sonar imagery, and acoustic backscatte r. Attention is given to choosing proper waveforms for particular appl ications. Using artificial neural networks (ANNs), evidence is fused f rom multiple-waveform types that detect local features. The wavelet wa veforms and their dilation and shift parameters are adaptively compute d with ANNs to maximize classification accuracy. Emphasis is placed on the acoustic backscatter case study, involving detecting a metallic m an-made object from natural and synthetic specular clutter with reverb eration noise. The synthetic clutter is shown to be a good model for t he natural clutter, which appears promising for avoiding huge data col lection efforts for natural clutter and for better delineating the cla ssification boundary. The classifier computes the locations, sizes, an d weights of Gaussian patches in time-scale space that contain the mos t discriminatory information. This new approach is shown to give highe r classification rates than an ANN with commonly used power spectral f eatures. The new approach also reduces the number of free parameters i n the classifier based on all wavelet features, which leads to simpler implementation for applications and to potentially better generalizat ion to test data.