The application of automated recognition techniques to side-scan sonar imagery

The rapid identification of ocean-bottom toxic-waste deposits has grown in importance due to worldwide industrialization and the expense of proper disposal. Acoustic surveys with a side-scan sonar provide a cost effective means of locating such damaging deposits. Here we describe an automatic image recognizer that can serve to prescreen volumes of data and alert a human operator. This recognizer operates using features and rules drawn from operator experience to establish candidate detections in amplitude-segmented imagery. The algorithm is described, performance examples are given, and applications are discussed.<
>