Measurement of Quantitative Characteristics of Schooling Movement of Fish by Computer Vision Methods

Abstract

In the study of the mechanisms of schooling behavior of fish, it becomes necessary to process large amounts of visual information. Early this processing was done mostly by hand. The paper study the possibility of automating the processing of observation results using the open CV computer vision package. Both individual frames and video fragments (scenes) of observations were processed. To process individual frames, a program was written for automatically counting the number of fishes in a school, another programs were written for determining the position of fishes and the degree of polarization of a group. During processing fragments of observations (scenes) the video scenes of fright of the school and video scenes of interaction between the predator and the school were considered. When studying the fright scene, a program for determining the speed of movement of individuals, a programs for calculating the kinetic energy of an individual moving and the kinetic energy of a group of fishes were built. The concept of the fright’s magnitude caused by a short local perturbation is introduced, and a method is proposed for numerically measuring the magnitude of the fright. When studying the scene of interaction between a predator and a school, a program for tracking the movement of an individual in a school was built. The proposed methods can be used in constructing and identifying the parameters of mathematical models of schooling fish movement.