This paper presents ORB-SLAM, a feature-based monocular SLAM system designed for real-time operation in both indoor and outdoor environments. The system is robust to severe motion clutter, supports wide baseline loop closing and relocalization, and includes full automatic initialization. ORB-SLAM uses the same features for all SLAM tasks: tracking, mapping, relocalization, and loop closing, making it efficient, simple, and reliable. The system employs a survival of the fittest strategy to select points and keyframes, generating a compact and trackable map that only grows if the scene content changes, enabling lifelong operation. Extensive evaluations on 27 sequences from popular datasets demonstrate ORB-SLAM's superior performance compared to other state-of-the-art monocular SLAM approaches. The source code is made publicly available to benefit the community. Key contributions include the use of ORB features for real-time performance, an improved place recognition method, and a novel initialization method based on model selection between homography and fundamental matrix. The system's robustness and accuracy are highlighted through experiments in challenging scenarios, such as the NewCollege robot sequence and the TUM RGB-D benchmark.This paper presents ORB-SLAM, a feature-based monocular SLAM system designed for real-time operation in both indoor and outdoor environments. The system is robust to severe motion clutter, supports wide baseline loop closing and relocalization, and includes full automatic initialization. ORB-SLAM uses the same features for all SLAM tasks: tracking, mapping, relocalization, and loop closing, making it efficient, simple, and reliable. The system employs a survival of the fittest strategy to select points and keyframes, generating a compact and trackable map that only grows if the scene content changes, enabling lifelong operation. Extensive evaluations on 27 sequences from popular datasets demonstrate ORB-SLAM's superior performance compared to other state-of-the-art monocular SLAM approaches. The source code is made publicly available to benefit the community. Key contributions include the use of ORB features for real-time performance, an improved place recognition method, and a novel initialization method based on model selection between homography and fundamental matrix. The system's robustness and accuracy are highlighted through experiments in challenging scenarios, such as the NewCollege robot sequence and the TUM RGB-D benchmark.