An underwater visual SLAM system with adaptive image enhancement

Underwater monocular visual simultaneous localization and mapping (SLAM) plays a crucial role in the navigation and localization of underwater robots. Low-light and turbid underwater environments pose significant challenges to the effectiveness and accuracy of these systems. This paper proposes a novel recognition algorithm based on the AquaVisNet model, designed specifically for such environments. Furthermore, an image enhancement algorithm tailored for these challenging environments is proposed that utilizes a serial-parallel fusion processing strategy. Such enhancement improves image quality significantly. Building on these advancements, an adaptive image enhancement ORB-SLAM (AIE-ORB-SLAM) system is presented for low-light and turbid underwater environments. The experimental results demonstrate that this system significantly outperforms the ORB-SLAM3 system in terms of various metrics. Under low-light, turbid, and combined conditions, the AIE-ORB-SLAM system improves the initialization time by 23.46%, 23.88%, and 81.69%, respectively; the tracking duration by 72.63%, 235.12%, and 294.29%, respectively; the number of keyframes by 74.71%, 140.00%, and 218.48%, respectively; the number of point clouds by 119.19%, 187.92%, and 317.11%, respectively; and the localization accuracy by 90.04%, 75.61%, and 66.81%, respectively. These results demonstrate that the proposed method significantly enhances the robustness and localization accuracy of underwater visual SLAM systems in low-light and turbid environments.