Fish target recognition and counting based on Dual-frequency Identification Sonar
(1 College of Marine Sciences, Shanghai Ocean University,
Shanghai Engineering Research Center of Estuarine and Oceanographic Mapping, Shanghai 201306, China;
2 Shanghai Ocean University, Marine Rance Engineering Technology Research Center, Shanghai 201306, China)
Abstract：Fish resource survey is a basic work of fishery management, and the investigation of fish resources using underwater acoustic detection technology has become a mainstream method. Aiming at the acoustic data collected by Dual-frequency Identification Sonar (DIDSON) in the reservoir, the acoustic data post-processing software Echoview and related algorithms were used to conduct fish identification and counting research. First, the Kovesi image denoising method and the background difference method were utilized to remove speckle noise and water background in acoustic images, then targets were identified according to the echo threshold, the α-β trajectory tracking algorithm was used to track moving targets to prevent double counting, and finally fish counting and length extraction were performed. The results show that, compared with munual counting, fish counting error of this method is within 10%, and the average error is 7.2%, with a very high statistical accuracy. Studies have proven that DIDSON can be used for fish identification and counting, and has a very broad application space for fish resource detection and management in shallow waters.