An accurate and efficient monocular mixed match SLAM

This paper proposed a Mixed Match SLAM embedded with two match methods, planar patch based and visual descriptor based. Planar patch based method eliminates the cost of feature extraction for motion estimation. Visual descriptor based method is more robust for wide baseline match and loop closing. We use a reasonable statistic criterion to determine keyframe for redundancy avoidance. An efficient outlier detection method is employed to pick out feature match outliers trapped in local minimum. Quantitative comparison experiments with the state-of-the-art ORB-SLAM2 show that the proposed method achieves better performance in large region dynamic scenes due to a less redundant keyframe network. It also provides better performance than ORB-SLAM2 in most static benchmark scenes with higher efficiency.