Orientation based multi-scale corner detection for mobile robot application

The accuracy and robustness of SLAM critically depend on the properties of feature points. This paper presents a robust corner detection method suitable for indoor mobile robot navigation. We combine the advantages of edge-based corner detection and intensity-based corner detection. Orientation field estimation followed by image-based curvature estimation can detect corners stably and accurately. Experimental results validate the robustness of the proposed corner detector in indoor environments.