An operational system for automatic school identification on multibeam sonar echoes

摘要

A system for identifying and tracking fish schools is demonstrated, based on the analysis of multibeam sonar data obtained by a Simrad SP90 long-range sonar. Fish-school detection and identification techniques are similar to those commonly used for vertical echosounders, further enhanced with innovative processing algorithms applied to successive multibeam echograms, increasing the certainty that the identified objects are fish schools. Additionally, analysis of school dynamic parameters facilitates the classification of targets into certain groups, here discriminating the fish aggregating device-natant fish complex from tuna. Statistical analysis of selected tracks quantifies the spatio-temporal variability of the school descriptors, which are used retrospectively to select appropriate analysis thresholds. The algorithms are implemented in an acquisition, visualization, and processing software platform that is flexible regarding sonar characteristics (beam width and number of beams) and can be extended easily to track school echotraces in a three-dimensional mode.