SLAM algorithm for indoor scene vision based on RGB-D cameras

Aiming at the problem of poor map construction and low localization accuracy of visual SLAM in complex scenes, this paper introduces a SLAM algorithm for indoor scenes based on RGB-D cameras, which utilizes sensor depth information and colour image information to achieve the construction and real-time localization of 3D maps of indoor scenes. In feature point extraction and information matching, the ORB feature point detection and matching algorithm is used to optimize the localization accuracy. It is demonstrated through experiments that a higher accuracy indoor scene building task is achieved with limited computational resources and provides support for path planning and autonomous navigation of mobile robots.