Abstract: In order to improve the accuracy of SLAM (simultaneous localization and mapping), a monocular SLAM method is proposed based on temporal-spatial consistency. Stable feature points usually possess the following two characters, that they appear in multiple continuous frames, and can be observed by cameras from different viewpoints. In this paper, these two characters are described by temporal consistency and spatial consistency, that is temporal-spatial consistency in short. The temporal consistency decides the timing of keyframe insertion, and the spatial consistency filters the 3D points strictly. On the KITTI dataset, the proposed method is compared with the ORB-SLAM (SLAM system based on oriented FAST and rotated BRIEF features) algorithm. Fewer keyframes need to be selected, so the keyframe poses are optimized more completely in the local optimization thread, and the processing speed is up to 35frames/s to meet the real-time performance requirements. Experiments show that the proposed method can effectively reduce errors and improve the accuracy of SLAM.