Automated Acoustic Method for Farmed Fish Counting and Sizing during Its Transfer Using DIDSON

摘要

Counting and sizing the farmed fish of large-size i.e. tuna, are often performed during the time of its transfer from one net cage to another. DIDSON provides an automated fish counting and sizing tool. However, neither its counter nor its sizer is suitable for measuring the farmed fish during its transfer because the transfer net is waved by currents and the fish image is often broken to several remains. This paper presents a fully automated acousuc method that can count and stze farmed fish during its transfer by using DIDSON image. The background is subtracted after being stabilized by a phase only correlation method. Tracing the edges using contour tracing method, the segmentation of fish is obtained. to prevent recounting the same fish. a biologically relevant Kalman filter algorithm was designed and adopted to predict the fish movement. Through analysis of the spatiotemporal trajectory of the track, the automated counting was performed. The broken fish images were searched, and the body length was obtained by summing down the center line segments from the head to the tail of fish. The proposed system was vexed to achieve the sizing accuracy within the range of -2.4 to 2.8 cm with the mean -0.4 cm by using the farmed YellowtaiL seriola quiwqusradiata(mean total length 83.1 cm).