Polarization-Enhanced Underwater Detection Method for Multiple Material Targets Based on Deep-Learning

Abstract

Underwater target detection is an essential topic in the applications of underwater exploration. However, underwater target detection faces serious challenges, such as complex scattering, low visibility, and target clutter. Here a polarization-enhanced underwater multiple material target detection method is proposed to address these challenges. The similarity principle of locally backscattered polarization features is utilized to suppress the influence of backscattered light. Our target detection model combines polarization gradient and edge detection techniques to optimize the detection process, enabling superior target detection and feature extraction. Experimental results indicate that our method has significantly enhanced the detection performance in multiple (overlapping or nonoverlapping) material targets, especially in high turbid underwater scattering environments. This research provides a promising new approach for polarized target detection in underwater environments and opens up new possibilities for underwater multiple-material target detection.