Underwater target detection and recognition based on cross-modal fusion of flow and electric information

Drawing inspiration from the natural sensing mechanisms of fish, this paper proposes, for the first time, a fusion method that combines the flow and electric modalities to perceive underwater moving targets through crossmodal information fusion, thereby overcoming the challenges of detection uncertainty and limited recognition accuracy. We present an array that can simultaneously measure the disturbance of both the flow and electric fields induced by the underwater target. Our approach introduces principal component analysis to enhance the robustness of detection and a dual-physics fusion algorithm that integrates tri-type artificial neural networks with Dempster-Shafer evidence theory to improve recognition accuracy. Experimental results show an 8% improvement over single-modal detection, achieving 97.5% recognition accuracy under varying conditions. This work provides a promising framework for leveraging cross-modal underwater information, significantly advancing target detection and recognition capabilities.