Elimination of Unnecessary Feature Points by Semantic Segmentation Aimed at Improving the Accuracy of Feature-Based SLAM

Abstract

In this research, we propose the method for improving the accuracy of feature-based SLAM. Instead of giving images directly to SLAM, partly blurred images to eliminate unnecessary feature points that could influence the accuracy of SLAM are given. To generate partly blurred images, we employ Semantic Segmentation that generates binary mask images to separate biological objects and surgical tools. By performing the image blur process only in the region of surgical tools, we eliminate feature points on the surgical tools that impair the accuracy of SLAM. We conducted preliminary experiments and confirmed our proposed method worked well in some situations.