Estimating the swimming speed of salmon in aquaculture sea cages using computer vision

Abstract

In a commercial aquaculture farm, knowledge about fish behaviour can influence farm management, fish growth and operating costs. In this presentation we describe research towards the development of the first stage of a system that will track fish movement in a sea cage and automatically analyse fish behaviour based on the movement. This first stage involves the design and implementation of a tracking system that will calculate the average speed and direction of fish in a sea cage. The system uses an underwater camera which captures video to a digital recording device. Captured video is then analysed, using computer vision software, to produce an average speed for every sampling period. The software can track up to 8 fish at a time and uses the Kalman Filter technique for estimation of fish position and the Global Nearest Neighbour algorithm for measurement-to-track association.