Abstract: In order to improve the ability of mobile robots to perform complex tasks in unknown environments, this paper constructs a visual localization and detection system that combines the concept of maximum information entropy with Simultaneous Localization and Mapping (SLAM). Based on the ORB-SLAM2 detection algorithm, under the premise of uniform distribution, the feature points with the most prioritized information under monocular vision are searched. The first N features with the maximum information entropy are then selected to be re-optimized for fast convergence to achieve high accuracy localization. At the same time, to verify the effectiveness of the algorithm, physical tests are conducted in combination with YOLO-V4 target detection, which can achieve real-time localization and detection in embedded mobile devices. The experimental results show that the localization accuracy of the proposed algorithm in this paper is improved in both TUM and KITTI datasets, and the proposed algorithm is better than the original algorithm in multiple scenarios and devices.