Wavelet Based Feature Extraction for Target Recognition and Minefield Detection

摘要

This project produced advances in the theory of wavelets and two- channel filter banks, and the development of new algorithms for the generation of wavelet filters and the wavelet based processing of image data, with a view towards their usefulness in image analysis for target recognition. These results include implementation of simulated annealing and Discrete Wavelet Transform algorithms, derivation of parameterizations for various useful spaces of wavelets, derivation of expressions for frequency and spatial uncertainty in wavelets, generation of wavelets optimized for different balances between spatial and frequency uncertainties, and development of wavelet transform domain denoising algorithms for feature detection algorithms. Much of the research was done on-site at the Naval Surface Warfare Center, Dahlgren, VA. Several collaborations were formed with NSWC scientists, and these produced accomplishments in addition to those in the grant proposal. Also, the P.I. presented tutorial courses and seminars to NSWC personnel. Some of the research was performed during visits to universities in South Africa, resulting in further useful and on-going collaborations. The grant supported a total of 6 graduate students (one Doctoral and 5 Masters) who performed software development and some theoretical derivations. During the period of the grant, 13 peer-reviewed papers were published (3 in journals and 10 at conferences).