An improved visual SLAM method based on image enhancement and adaptive threshold

Visual simultaneous localization and mapping (VSLAM) systems have advanced significantly in autonomous driving. However, traditional VSLAM algorithms face challenges such as fixed threshold selection, error extraction, and mismatches in low-light or low-texture environments. This paper presents an improved VSLAM algorithm designed to enhance the accuracy and robustness of feature point extraction, leading to more reliable map construction and positioning. Experimental results demonstrate that the enhanced VSLAM algorithm improves localization accuracy and robustness across various textures and complex lighting conditions, showcasing its practical application value.