FAST and FLANN for feature matching based on SURF

The aim of this paper is to Reduce Speeded Up Robust Features' (SURF) time-consuming problem and get a high accuracy in image registration, the Features From Accelerated Segment Test (FAST) algorithm and the Fast Library for Approximate N earest Neighbors (FLANN) are proposed respectively for extracting the feature points and increasing the accuracy. First, the FAST algorithm is applied to extract the features of the image, and then the SURF algorithm is used to construct the descriptor to realize the rapid extraction of image features with rotation invariance. And then an improved feature matching algorithm FLANN is proposed to accurately match the feature points. The experimental results show that our method is about two times faster than the traditional SURF algorithm and also has a higher accuracy. The main outcomes of this paper are the usage of FAST for a time-consuming problem and FLANN for a high accuracy in image registration.