A Fast Feature Extraction Process for Visual SLAM

Abstract

Over the past decades, visual SLAM has successfully applied in robotics and augmented reality. The effectiveness of the feature extraction has an important influence on the performance of the visual SLAM. This paper proposes an Oriented AGAST and Rotated BRIEF (OARB) method to improve the efficiency of visual SLAM to address the specific application, such as mobile platform. We use the AGAST algorithm to detect corner points in parallel and measure the direction of each corner. Then we use the BRIEF algorithm to calculate the descriptor. We compare our proposed OARB method with the ORB method in visual SLAM on two public datasets. Experimental results demonstrate that our proposed OARB method can outperform the ORB method for visual SLAM in terms of speed and meanwhile achieve the competitive performance.