Robust Feature Matching Based on Adaptive ORB for Vision-based Robot Navigation

Feature matching technology is one important part for a vision navigation system to estimate robot pose according to natural landmarks. However, it is difficult for the existing techniques to solve the dilemma between accuracy, and efficiency, as well as the robustness under a complex workspace involving non-uniform illumination and texture-less. To address this issue, this paper proposes a novel feature matching algorithm that is robust, accuracy, and relatively fast in complex workspaces. The main idea of the proposed method is a multi-stage matching module consisting of K Nearest Neighbor, Threshold Filtering, Eigenvector Norm, and Random Sample Consensus (KNN-TF-EN-RANSAC) to eliminate mismatches for retaining the optimal matches. Meanwhile, the proposed matching module is computed on adaptive Oriented fast and Rotated Brief (ORB) features extracted by a variable extraction radius. Finally, extensive experiments demonstrate the robustness, accuracy, and real-time performance of the proposed method.