Abstract: This work focuses on coherence-based methods for multi-aspect detection and classification of underwater targets using broadband sonar. Using multiple sonar pings or aspects of an object is an excellent way to improve classification rates and increase the confidence in the decisions made. The new frequency subband coherence analysis feature extraction method developed in this work uses multiple sonar returns off an object, where each sonar return is characterized by its specific frequency subbands. The coherence patterns extracted from the most discriminatory frequency subbands of two sonar returns off mine-like objects are found to be different than those of non-mine-like objects, and hence provide excellent discrimination using only a simple classifier. This method offers a more rigorous way of performing acoustic color processing, where ping-to-ping coherence between sonar returns is also exploited in extracting acoustic color features.<br>
A new method of performing coherence-based blind synthetic aperture sonar (SAS) processing is also introduced. SAS processing is an excellent tool to accompany the multi-aspect feature extraction and classification methods used in this study, since it can generate images that accurately represent the contents of the seafloor for object detection and localization. The new coherence-based SAS processing method alleviates many of the major drawbacks of traditional SAS processing, since it does not require elaborate platform motion estimation and compensation techniques. Moreover, the new coherence-based method produces images that covey information that is useful not only for underwater object detection and localization, but also object classification. A comprehensive study is carried out to compare the performance of the new coherence-based feature extraction and SAS processing methods with previously implemented methods. Data used in this study consists of sonar pings from a mixture of mine-like and non-mine-like objects collected in two different target fields under a variety of operating and environmental conditions.
Adviser: Dr. Mahmood R. Azimi-Sadjadi Co-Adviser: N/A Non-ECE Member: F. Jay Breidt (Statistics) Member 3: J. Rockey Luo Addional Members: N/A