Author(s): Xiaotao Shi; Jiawei Xu; Zhiyong Huang; Christos Katopodis; Xuan Ban; Boyd Kynard; Junjun Tan; Haifeng Zhu; Chenyu Lin

Linked Author(s):

Keywords: Automatic tracking; computer vision; fish trajectories; flow–fish interface; fuzzy c-means clustering

Abstract: ABSTRACTEcohydraulics research is just beginning to have all the tools needed to examine the 2-dimensional velocity preference of fish in flow fields. We developed an experimental system with a gradient flow field in a test channel to observe flow-field preference of juvenile silver carp, Hypophthalmichthys molitris. We used automatic methods, which acquire fish swimming tracks using computer vision (video). We examined the flow field using hydrodynamic software simulation; then, using the fuzzy c-means clustering algorithm, each frame of the flow field was divided into areas of different velocity. Finally, the automatic fish trajectory tracking was coupled with hydraulic simulation of the flow field to compute an estimate of the time fish spent in each velocity area. To validate results from the automatic method, we used manual examination of videos of swimming juvenile silver carp to analyse their preferred velocity fields. Results show the automatic method greatly reduces processing time compared to the manual method.

DOI: https://doi.org/10.1080/00221686.2018.1538020

Year: 2019