Features detection and matching for visual simultaneous localization and mapping (VSLAM)

This paper presents the feature detection method for aerial image. The image captured from the navigation was used to select the best landmarks for localization and mapping in SLAM. A robust visual detection method has contributed to better landmark and data association selection. Therefore, different feature detection algorithms were compared to evaluate the best landmark detector and descriptor for the VSLAM. The performances of the feature detectors were evaluated using dataset provided by the Robotics Research Group at University of Oxford. The local images of matching effect on the detector and descriptor have proved the correctness of key point matching. The selected method has been validated and proven efficient for the VSLAM.