A simple approach to count and track underwater fishes from videos

Abstract

Fishes are of great importance to the ecosystem. Behavior of fishes is interesting. Counting and tracking of fishes can provide good knowledge about the behavior of fishes. Counting and behavior quantifying of fishes within a turbulence or trawl environment are challenging tasks. The traditional methods are not only inefficient but also expensive. Thus counting and tracking under water fishes from videos are emerging topic for ichthyologists. This paper addresses a simple method to count and track underwater fishes from videos. It is a hybrid of background subtraction, Hungarian algorithm, and Kalman filter. It enables tracking of fishes whose number can vary over time. Theoretical runtime of the tracking algorithm is O(n(3)) with problem size n. Experimental results demonstrate its effectiveness.