A Novel Monocular SLAM Algorithm for High Real-Time Based on Kalman Filter

Abstract

Simultaneous Localization and Mapping (SLAM) has been a hot research direction in the field of mobile robots since it was proposed. ORB-SLAM is a typical algorithm in feature point SLAM system. However, because the ORB-SLAM can only provide sparse map construction, it can only be used for positioning. Just using ORB-SLAM to handle the positioning problem make the calculation complexity higher and reduce the real-time performance. Because of the complicated calculation, the ORB-SLAM is not practical to the embedded devices with low computing power. Therefore, we propose a monocular SLAM algorithm with high real-time performance based on Kalman filter. Our proposed algorithm reduces the time consumption of the system by modifying the posture optimization algorithm. Our algorithm provides accurate and efficient positioning for upper-level applications. We use Kalman filter modeling to optimize the pose of the camera. The experimental results show that compared with traditional algorithm, our proposed algorithm can effectively improve the real-time performance and reduces the time consumption of the system.