Experimental Study on the Acoustic Scattering Characteristics and Classification of Typical Freshwater Fish and Crustacean Species

Abstract

This study investigates the acoustic scattering characteristics and classification of four freshwater species (Channa argus, Oreochromis niloticus, Homarus americanus, and Pelodiscus sinensis) using neural networks. Standard underwater acoustic measurements yield full-aspect horizontal scattering data from controlled tank experiments. Time-domain analysis extracts key echo features including highlight count, amplitude, time-delay differences, and pulse-width broadening, while Radon transform imaging reveals structure-scattering correlations. Frequency domain reveals interspecies differences in acoustic target strength by analyzing frequency-dependent scattering characteristics across different frequencies and incident angles. Statistical analysis demonstrates that the target strength distributions of the four freshwater species generally follow \(\chi^{2}\) patterns. Finally, we propose a classification method based on time–frequency-domain acoustic scattering characteristics of biological targets. A backpropagation neural network (BPNN) model incorporating these time–frequency-domain scattering characteristics achieves 95% classification accuracy. This study conducts neural network classification research based on multidimensional acoustic scattering characteristics of aquatic biological targets, extending the applications of acoustic technology in fisheries exploration and aquaculture industries. The work will provide new methodological insights for deep integration of neural networks with aquaculture practices.