Classification of pelagic fish using wideband echosounders

Abstract

The landing obligation for pelagic fisheries was introduced in January 2015 by the European Commission and will gradually be extended to all fisheries in 2019. This regulation follows the trend for sustainable fisheries in Europe and states that all catches must be landed and counted against the quotas. Pre-catch acoustic fish classification can remotely provide information about fish school composition in terms of species and size distribution. This technique has been tested using three commercial wideband transceivers mounted on a commercial freezer trawler. Together, the three SIMRAD EK80 transceivers cover a frequency range of 45 kHz to 260 kHz. Acoustical data was collected for four months during fishing operations in the North Sea and the English Channel. Custom, through-the-sensor image processing algorithms have been developed to perform automatic school detection under severe noise, interference and ship motion conditions. Dynamic Factor Analysis was applied to exploit common trends in the spectra of the species, resulting in classification scores of 95-100% for the set of 38 homogeneous schools of herring, mackerel and horse mackerel. This data set is too limited to draw firm conclusions on potential operational performance. However, since the utilized spectral trends match with theoretical expectation, the approach is promising.