Accuracy Evaluation of Stereo Visual SLAM with Unnecessary Feature Point Elimination Using Blur Processing for AR Surgical Support System

In this study, we describe an evaluation of the positional accuracy of stereo Visual SLAM with noisy feature points removed by blurring. We constructed an experimental environment that intentionally degrades the accuracy of Visual SLAM. In this environment, we placed the objects that are likely to generate noisy feature points. We evaluated the accuracy of the estimated positions of the original video and the video in which blurred regions of noise in the original video were provided to Visual SLAM. Consequently, we confirmed that the accuracy was improved compared to that of the original video.