Research on Feature Matching of an Improved ORB Algorithm
Liming Zhao, Jingfan Yang, Yi Zhang, Junzhi Huang · 2022 IEEE 6th Information Technology and Mechatronics Engineering Conference (ITOEC) · 2022
Aiming at the problems of ORB (Oriented FAST and Rotated BRIEF) algorithm in complex environments, such as low matching accuracy and poor real-time performance. An improved AGAST feature point detection fusion ORB algorithm is proposed. First, build Gaussian pyramid to generate scale-invariant AGAST, then use gray-scale centroid method to generate directional feature point descriptors, and finally use Hamming distance to complete the matching, and combine RANSAC to optimize the matching results; at the same time, this article focuses on the AGAST algorithm The fixed threshold is improved in the middle, and an adaptive threshold method is proposed. The experimental results show that, compared with the original ORB algorithm, the algorithm proposed in this paper improves the time by 15.41%, and has a good matching accuracy under the environment of varying illumination and rotation scale.