Fish detection automation from ARIS and DIDSON SONAR data

Abstract

The goal of this thesis is to analyse SONAR files produced by ARIS and DIDSON manufactured by Sound Metrics Co. which are ultrasonic, monostatic and multibeam echo-sounders. They are used to capture the behaviour of Atlantic salmon, which recently has been on the lists of endangered species. These SONARs can work in dark lighting conditions and provide high resolution images due to their high frequencies that ranges from 1.1 MHz to 1.8 MHz. The thesis goes through extracting data from file, redrawing it, and visualising it in human friendly format. Next, images are analysed to search for fish. Results of analysis are saved in formats such as JSON, to allow harmony with other legacy systems. Also the output helps in future development due to the support for JSON in multitude of programming languages. Eventually, a user-friendly user interface is introduced, which helps making the process easier. The software is tested against data-sets from rivers in Finland, that are rich in Atlantic salmon.