Image Feature Point Extraction and Matching Method Based on Mobile Augmented Reality

Abstract

In the mobile augmented reality system, the extraction of image feature points, the calculation of descriptor and the matching of feature points are the foundation and key of 3D tracking registration technology. In order to meet the real - time and stability of the mobile augmented reality system, improve the robustness of the algorithm, and realize the accuracy of the virtual real fusion of the augmented reality system, the image feature point extraction and matching method based on the mobile augmented reality technology is studied. This paper proposes an improved image feature point extraction and matching algorithm. S-ORB (Scale ORB) algorithm is proposed. The image pyramid is constructed by three times Gaussian filtering and three times down sampling. Fast corner is used to extract image feature points. The feature points are located by interpolation and described by rotation related brief sub calculation. Whether the feature descriptors match or not is calculated by hamming distance. RANSAC algorithm is used to filter and extract image feature points The false matching points are eliminated and the feature points are extracted and matched based on mobile augmented reality. The experimental results show that the accuracy of the improved S-ORB algorithm in the process of image feature point extraction and matching is significantly higher than that of ORB algorithm and brisk algorithm, and it has faster calculation speed. It not only improves the accuracy of image matching. It has better robustness, and better meets the real-time requirements of mobile augmented reality system.