Solar Image Matching Based on Improved Freak Algorithm

FREAK algorithm has the defects of not having scale invariance and single feature point matching strategy, being prone to unsatisfactory results. Based on AKAZE and RANSAC algorithm, an improved FREAK algorithm is proposed, named AKAZE-FREAK. The image pyramid was constructed by using the nonlinear...

Full description

Saved in:
Bibliographic Details
Main Authors: SONG, MEI-PING, CAO, YUE-JING, YU, CHUN-YAN, AN, JU-BAI, CHANG, CHEIN-I
Format: Conference Proceeding
Language:English
Subjects:
AKAZE
Approximation algorithms
Brightness
Feature extraction
Feature point matching
Filtering algorithms
FREAK
Image matching
Improvement FREAK
Lighting
Matched filters
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:FREAK algorithm has the defects of not having scale invariance and single feature point matching strategy, being prone to unsatisfactory results. Based on AKAZE and RANSAC algorithm, an improved FREAK algorithm is proposed, named AKAZE-FREAK. The image pyramid was constructed by using the nonlinear diffusion filter and the numerical solution was obtained by fast display diffusion (FED) to obtain the image point coordinates with sub-pixel precision. Then the feature points were described by the FREAK descriptors to distribute directions for the feature points. Finally, based on the initial matching of the hamming distance to the eigen vector, RANSAC is used to eliminate the error matching point. In this paper, this algorithm is applied to the recognition of repeated shooting solar images. After the improved algorithm is compared with the SIFT-FREAK, SURF-FREAK and FREAK algorithms, it shows an improvement in the image feature point matching accuracy with good robustness on the scale difference, illumination difference and rotation difference of the image. It also improves the accuracy of the judgement of the late repeat.
ISSN:2160-1348
DOI:10.1109/ICMLC.2018.8527012