An Automatic Registration Method of Remote Sensing Imagery Based on FAST Corner and SURF Descriptor

Abstract

An automatic geometry registration method based on Features from Accelerated Segment Test(FAST) corner detector and Speeded Up Robust Features(SURF) is proposed in this paper.Firstly,applying HSI transform on both the reference image and the image to registration,and then building gauss pyramid of the images.Secondly,detecting and extracting the FAST corner points of both images,and calculating the SURF descriptors of the corner points,following by searching match point pairs by K-D tree.Thirdly,iteratively using partial minimum least squares to remove error point pairs and then calculate the geometry transform coefficients.Lastly,excuting the geometry transform to get the registration image.An experiment on two groups of images was performed,in which the proposed method was respectively compared with the automatic registration method based on SURF features and the method used in ENVI software to obtain ground control points automatically,and the results show that the proposed method can get more match points and obtain higher geometric accuracy,except which is slightly inferior to SURF algorithm in scale invariance.