计算机科学 ›› 2015, Vol. 42 ›› Issue (9): 29-32.doi: 10.11896/j.issn.1002-137X.2015.09.006
郝昀超,王显珉
HAO Yun-chao and WANG Xian-min
摘要: 基于SIFT算法的遥感图像配准精度高、稳定性强,但图像幅宽大、提取特征点数量多使得配准过程耗时长。提出了一种高分辨率遥感图像配准的并行加速方法。该方法在特征点提取时利用GPU实现了高斯金字塔建立过程中的并行加速,并对提取出的大量特征点使用共享内存来进行局部极值高速缓存,降低了特征点提取所需的运算时间;同时通过分块处理以及OpenMP多线程技术实现了特征点匹配及仿射模型计算过程的CPU并行处理。实验表明:本方法相对于传统的SIFT算法平均加速3倍,并且对于固定大小的图像,本方法的特征点提取时间和特征点个数具有线性关系,加速比随着提取出特征点数量的增加而增大。
[1] Zitova B,Flusser J.Image registration methods:a survey[J].Image and Vision Computing,2003,21:997-1000 [2] Li Qiao-liang,Wang Guo-you,Liu Jian-guo.Robust scale-inva-riant feature matching for remote sensing image registration[J].IEEE Geoscience and Remote Sensing Letters,2009,6(2):187-291 [3] Zhang Yun-sheng,Zhou Pei-long,Ren Yue,et al.GPU-accele-rated large-size VHR images registration via coarse-to-fine mat-ching[J].Computers and Geosciences,2014,66:54-65 [4] 雷小群,李芳芳,肖本林.一种基于改进SIFT算法的遥感影像配准方法[J].测绘科学,2010,35(3):143-145 Lei Xiao-qun,Li Fang-fang,Xiao Ben-lin.A registration method of RS image based on improved SIFT algorithm[J].Science of Surveying and Mapping,2010,35(3):143-145 [5] Kirk D B,Wen-mei W.Programming massively parallel processors:a hands-on approach[M].Morgan Kaufmann,2010 [6] Dagum L,Menon R.OpenMP:an industry standard API forshared-memory programming[J].Computational Science & Engineering,1998,5(1):46-55 [7] Lowe D G.Distinctive image features from scale-invariant keypoints[J].International Journal of Computer Vision,2004,60(2):91-110 [8] CUDA C Programming Guide.http:docs.nvidia.com/cuda/cuda-C-programming-guide/#axzz3iTPutLEx [9] Nvidia CUDA Computer Unified Device Architecture[S].Programing Guide,Version 2.0 beta 2,8 [10] 周海芳,赵进.基于GPU的遥感图像配准并行程序设计与存储优化[J].计算机研究与发展,2012,9(S1):281-286Zhou Hai-fang,Zhao Jin.Parallel programming design and stora-ge optimization of remote sensing image registration based on GPU[J].Journal of Computer Research and Development,2012,49(S1):281-286 |
No related articles found! |
|