Benchmarking and Comparing Popular Visual SLAM Algorithms
share
This paper contains the performance analysis and benchmarking of two popular visual SLAM Algorithms: RGBD-SLAM and RTABMap. The dataset used for the analysis is the TUM RGBD Dataset from the Computer Vision Group at TUM. The dataset selected has a large set of image sequences from a Microsoft Kinect RGB-D sensor with highly accurate and time-synchronized ground truth camera poses from a motion capture system. The sequences selected includes variety of scenes and camera motions which are ideal for testing the robustness of SLAM algorithms in different situations. The evaluation metrics used for the comparison are Absolute Trajectory Error (ATE) and Relative Pose Error (RPE). The analysis involves comparing the Root Mean Square Error (RMSE) of the two metrics and the processing time for each algorithm. The test sequences are intentionally selected to depict variety of problems and camera motions faced by SLAM algorithms. This paper serves as an important aid for selection of SLAM method for different scenes and camera motions. The analysis helps to realize the limitations of both the SLAM methods. The paper also points out some underlying flaws in the used evaluation metrics.
READ FULL TEXT