一种改进AKAZE特征和RANSAC的图像拼接算法

doi:10.19678/j.issn.1000-3428.0056874

摘要/Abstract

摘要： 针对传统图像描述方法在图像对变化复杂时特征点配准精度低，且传统RANSAC算法计算稳定性差的问题，提出一种结合改进AKAZE特征与RANSAC算法的图像拼接算法。利用AKAZE算法构造非线性尺度空间提取图像特征点，采用卷积神经网络描述符生成128维特征向量描述图像特征点，通过精简特征点并在迭代中设定嵌套阈值改进RANSAC算法得到最优变换矩阵模型，结合最佳缝合线算法和多频段融合算法对变换后的图像进行拼接。实验结果表明，和传统AKAZE算法相比，该算法在图像对的视角差异和光照差异较大时，配准精度分别提高12.60和6.99个百分点，改进后的RANSAC算法计算时间较改进前缩短4.17 ms，图像拼接精度更高。

关键词: 图像拼接, 图像配准, AKAZE算法, 卷积神经网络描述符, RANSAC算法

Abstract: In view of the low registration accuracy of the traditional image description methods for feature points in the case of complex image pair changes,and the traditional RANSAC algorithm has poor computational stability,this paper proposes an image stitching algorithm that combines improved AKAZE features and RANSAC algorithm.The AKAZE algorithm is used to construct nonlinear scale space to extract image feature points.The Convolutional Neural Network(CNN) descriptor is used to generate 128 dimensional feature vector to describe image feature points.The optimal transformation matrix model is obtained by simplifying the feature points and setting nested threshold during iteration to improve the RANSAC algorithm.The transformed images are stitched together by using the optimal stitching linear algorithm and multi-band fusion algorithm.Experimental results show that compared with the traditional AKAZE algorithm,the proposed algorithm increases the registration accuracy by 12.60 percentage points and 6.99 percentage points respectively in the cases of the large view angle difference and illumination difference.The calculation time of the improved RANSAC algorithm is reduced by 4.17 ms,and the image stitching accuracy is higher.

Key words: image stitching, image registration, AKAZE algorithm, Convolutional Neural Network(CNN) descriptor, RANSAC algorithm

中图分类号:

TP391

吴禄慎, 陈小杜. 一种改进AKAZE特征和RANSAC的图像拼接算法[J]. 计算机工程, 2021, 47(1): 246-254.

WU Lushen, CHEN Xiaodu. An Image Stitching Algorithm Based on Improved AKAZE Feature and RANSAC[J]. Computer Engineering, 2021, 47(1): 246-254.

https://www.ecice06.com/CN/Y2021/V47/I1/246

图/表 15

20210125164732

20210125164735

20210125164738

20210125164744

20210125164747

20210125164751

20210125164755

20210125164758

20210125164801

20210125164804

20210125164807

20210125164810

20210125164813

20210125164819

20210125164823

参考文献

[1] ZHANG Yi.Research on feature matching algorithm of UAV remote sensing image[D].Zhengzhou:PLA University of Information Engineering,2015.(in Chinese)张一.无人机遥感影像点特征匹配算法研究[D].郑州:解放军信息工程大学,2015.
[2] ZHAO Fuqun.Overview of image registration technology[J].Digital Technology and Application,2016,34(6):245-246.(in Chinese)赵夫群.图像配准技术研究综述[J].数字技术与应用,2016,34(6):245-246.
[3] XU Min,MO Dongming.Application of Canny edge feature 18 dimensional descriptor in image stitching[J].Computer Engineering,2017,43(9):310-315.(in Chinese)徐敏,莫东鸣.Canny边缘特征18维描述符在图像拼接中的应用[J].计算机工程,2017,43(9):310-315.
[4] RUBLEE E,RABAUD V,KONOLIGE K,et al.ORB:an efficient alternative to SIFT or SURF[C]//Proceedings of 2011 IEEE International Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2011:1-8.
[5] LI Yi,FENG Nan,TAN Shuncheng.Augmented reality method based on KCF and improved ORB[J].Computer Engineering,2019,45(8):230-235.(in Chinese)李一,冯楠,谭顺成.基于KCF与改进ORB的增强现实方法[J].计算机工程,2019,45(8):230-235.
[6] ALCANTARILLA P F,BARTOLI A,DAVISON A J.KAZE features[C]//Proceedings of 2012 European Conference on Computer Vision.Berlin,Germany:Springer,2012:52-58.
[7] ALCANTARILLA P F.Fast explicit diffusion for accelerated features in nonlinear scale spaces[C]//Proceedings of 2013 British Machine Vision Conference.Berlin,Germany:Springer,2013:21-27.
[8] CHEN Lingying,YANG Shiwu.Image stitching based on multi-algorithm combination in railway video monitoring[J].Journal of Electronic Measurement and Instrument,2012,26(3):229-235.(in Chinese)陈凌颖,杨世武.铁路视频监控中基于多算法结合的图像拼接[J].电子测量与仪器学报,2012,26(3):229-235.
[9] SONG Wei,WANG Yongbo,ZHANG Peipei.Research on UAV image stitching algorithm based on improved AKAZE features[J].Computer Engineering and Science,2019,41(5):873-878.(in Chinese)宋伟,王永波,张培佩.基于改进AKAZE特征的无人机影像拼接算法研究[J].计算机工程与科学,2019,41(5):873-878.
[10] SHEN Xueli,CHEN Xintong.Feature matching algorithm of triplet descriptor[J].Computer Application and Software,2019,36(2):205-210.(in Chinese)沈学利,陈鑫彤.三元组描述符的特征匹配算法[J].计算机应用与软件,2019,36(2):205-210.
[11] WANG Xiaohong,HE Zhiwei,DENG Shixiong,et al.Research on UAV remote sensing image matching using feature combination detection algorithm[J].Mapping Bulletin,2019,65(1):29-33.(in Chinese)王晓红,何志伟,邓仕雄,等.利用特征组合检测算法的无人机遥感影像匹配研究[J].测绘通报,2019,65(1):29-33.
[12] LI Jiahui,ZHANG Fengshou,CUI Haoyang.A homography matrix estimation method based on improved RANSAC method[J].Software Guide,2020,19(2):149-152.(in Chinese)李嘉惠,张丰收,崔浩阳.一种改进RANSAC算法的单应性矩阵估计方法[J].软件导刊,2020,19(2):149-152.
[13] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[14] GREWENIG S,WEICKERT J,BRUHN A.From box filtering to fast explicit diffusion[EB/OL].[2019-10-26].https://link.springer.com/chapter/10.1007/978-3-642-15986-2_54.
[15] LI F F,PERONA P.A Bayesian hierarchical model for learning natural scene categories[C]//Proceedings of 2005 IEEE Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2005:524-531.
[16] TIAN Yurun,FAN Bin,WU Fuchao.L2-Net:deep learning of discriminative patch descriptor in euclidean space[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2017:32-36.
[17] SHEN Tianwei,LUO Zixin,ZHOU Lei,et al.Matchable image retrieval by learning form surface[C]//Proceedings of 2018 Asian Conference on Computer Vision.Berlin,Germany:Springer,2018:415-431.
[18] LUO Zixin,SHEN Tianwei,ZHOU Lei.GeoDesc:learning local descriptors by integrating geometry constraints[C]//Proceedings of 2018 European Conference on Computer Vision.Berlin,Germany:Springer,2018:170-185.
[19] ZHANG Meiyu,WANG Yangyang,WU Liangwu,et al.Image mosaic algorithms based on DBSCAN clustering and mutual Information[J].Journal of Chinese Computer Systems,2020,41(4):825-829.(in Chinese)张美玉,王洋洋,吴良武,等.结合DBSCAN聚类与互信息的图像拼接算法[J].小型微型计算机系统,2020,41(4):825-829.
[20] FAN Yanguo,CHAI Jianglong,XU Mingming,et al.Image registration based on ORB and RANSAC fusion improvement[J].Optical Precision Engineering,2019,27(3):702-717.(in Chinese)樊彦国,柴江龙,许明明,等.基于ORB与RANSAC融合改进的图像配准[J].光学精密工程,2019,27(3):702-717.
[21] BROWN M,LOWE D G.Automatic panoramic image stitching using invariant features[J].International Journal of Computer Vision,2007,74(12):59-73.

选择文件类型/文献管理软件名称

选择包含的内容