AUTOMATIC INTERPRETATION OF SONAR IMAGE SEQUENCES USING TEMPORAL FEATURE MEASURES

This paper reports the development of a system for the automated inter pretation of sector scan sonar data. It proposes the use of a new comb ination of feature measures derived from sequences of sonar scans to c haracterize the behavior of targets' returns over time. Previous resea rch used grey-scale and shape descriptors derived from single sonar sc ans. However, problems were experienced with targets whose return vari ed significantly over time (such as divers, UUV's, and ships' wakes). Hence a new set of temporal feature measures has been developed by com bining existing one-dimensional temporal measures and two-dimensional object descriptors. These new features provide a quantitative descript ion of the behavior of a target's two-dimensional returns over a seque nce of sonar scans. Experiments with a limited but real data set have shown that classification accuracy can be significantly improved by th e use of these new features. The use of ''static'' feature measures (d erived from a single scan) was observed to give classification errors of between 7% and 10% when they were applied to the data set. In contr ast, the use of temporal measures reduced this error rate to 1% or 2% and in some cases reduced it to zero.