Advanced ORB-SLAM2 Based on Target Enhancement and Binary Transformation

Visual 3D reconstruction aims to extract detailed information of relevant real-world scenarios with the help of computer vision and sensor devices such as industrial cameras. Vision SLAM refers to the use of vision sensors to achieve self-positioning and perception of the environment. For indoor environments, RGBD camera provides rich visual and depth information to realize positioning and mapping of mobile robots. However, in the actual operation process, there are problems such as the lack of matching feature points when the lighting conditions are insufficient and low operating efficiency. In this paper, Azure Kinect combined with ORB-SLAM2 is utilized to build the system. The SLAM system is improved by pre-processing the input image and changing the reading mode of the ORB dictionary, which alleviates the above mentioned problems. The experiments show that the proposed methods improve the performance of the system and satisfactory SLAM mapping effect is obtained.