An Underwater Target Classification Method based on Optimized Integral Dictionary

The classification and identification of underwater targets holds a significant position in the domain of active underwater acoustic detection. Presently, target classification approaches based on machine learning and deep learning encounter numerous challenges when implemented on underwater unmanned platforms, and their applicability is constrained. Hence, this paper innovatively puts forward a target classification method grounded on sparse representation, which integrates the merits of integral dictionaries and dictionary learning. Specifically, the integral dictionary technology effectively lowers the dimension of the target waveform dictionary, thereby enhancing the execution efficiency of the algorithm; concurrently, the dictionary learning technology further boosts the sparsity of the integral dictionary, thereby escalating the classification accuracy and noise robustness capability. In the experimental validation, tests were carried out in an anechoic water tank. The outcomes reveal that under low signal-to-noise ratio conditions (-15 dB), the average time consumption of a single experiment of this method amounts to merely 3.21ms, and the classification accuracy reaches as high as 87.60%, demonstrating excellent application potential and broad prospects, which aligns with the requirements of underwater unmanned platforms.