Benthic fish behavior characterization with a mechanically scanned imaging sonar

Abstract

A point stationary, acoustic-based surveying system was proposed to fulfill the necessities of benthic fish behavior assessment. The surveying system can be split into two major and inter-dependent techniques, i.e., acoustic image acquisition and image post-processing. The first part comprises a high-frequency, mechanically scanned imaging sonar (MSIS) with bottom-fixed, side-looking working configuration. Major modules of the proposed image processing procedure include: stationary objects subtraction, region and textural feature extraction, unsupervised classification, fish target identification and quantification. For the specific case study conducted in a deep water fishpond, it is evident that both individual and school fish could be discriminated by image frames collected at a randomly selected point with scanning range setting at 5 m and frame rate at 75 s. Based on spatial and temporal analyses on position and area of the discriminated fish targets, it was concluded that fish movement pattern in the scanned area followed two distinctive corridors with significant different passage rate, i.e., a discrete, periodic and high passage rate mode in Corridor#1 and a continuous, steady and low passage rate mode in Corridor#2. Environmental features, such as bank slope, substrate mound and float raft represented specific meeting points for dynamic aggregations and schooling. Fish tended to converge into prominent schools subsequent to interactions with these features. The proposed system represented a practical and cost effective tool in acquiring image frames with sufficient spatial and temporal resolution for the characterization of fish behavior.