Optical Fish Trajectory Measurement in Fishways through Computer Vision and Artificial Neural Networks

Abstract

Vertical slot fishways are hydraulic structures that allow the upstream migration of fish through obstructions in rivers. The appropriate design of a vertical slot fishway depends on the interplay between hydraulic and biological variables because the hydrodynamic properties of the fishway must match the requirements of the fish species for which it is intended. One of the primary difficulties associated with studies of real fish behavior in fishway models is that the existing mechanisms to measure the behavior of the fish in these assays, such as direct observation or placement of sensors on the specimens, are impractical or unduly affect the animal behavior. This paper proposes a new procedure for measuring the behavior of the fish. The proposed technique uses artificial neural networks and computer vision techniques to analyze images obtained from the assays by means of a camera system designed for fishway integration. It is expected that this technique will provide detailed information about the fish behavior, and it will help to improve fish passage devices, which is currently a subject of interest in the area of civil engineering. A series of assays has been performed to validate this new approach in a full-scale fishway model with living fish. We have obtained very promising results that allow accurate reconstruction of the movements of the fish within the fishway.