基于异构多核构架的双目散斑3维重建
|
摘要
为提高双目立体匹配的速度和精度,提出一种基于异构多核构架平台实现的双目立体散斑3维重建方案。在构建双目立体视觉模型基础上,辅助投射白色散斑结构光,然后采集散斑变形图像并进行极线校正。对传统的ZNCC快速计算方法进行改进,运用零均值归一化互相关函数(ZNCC)作为相关算法的匹配代价函数,克服了传统立体视觉算法对弱纹理区域重建效果较差的缺点。将该算法移植到异构多核处理器Myriad2上,实现了物体快速高精度3维重建。实验结果表明借助异构多核构架处理器强劲的并行运行能力,在不损失系统重建精度的前提下,使系统运行时间大大缩短,对系统的重建效率具有较大提升。
In order to improve speed and precision of binocular stereo matching,this paper deals with 3D reconstruction with binocular speckle images based on heterogeneous multi-core processor.On the basis of passive binocular stereo vision model and with the assistant of projecting white binary speckle pattern,the deformed speckle images were captured and then polar-rectified.Exploiting the improved traditional fast-ZNCC calculation method and zero-mean normalized cross correlation(ZNCC) as the correlation function,the shortcomings of low precision and low efficiency of the conventional stereo vision algorithm for weak texture regional reconstruction were effectively overcame.The developed algorithm was implemented on Myriad2 platforms characterized by heterogeneous multi-core architecture achieving accurate and efficient three dimension reconstruction.The experiments demonstrated that the efficiency and performance in 3D reconstruction were significantly improved without any loss of accuracy assisted by the heterogeneous computational architecture.
In order to improve speed and precision of binocular stereo matching,this paper deals with 3D reconstruction with binocular speckle images based on heterogeneous multi-core processor.On the basis of passive binocular stereo vision model and with the assistant of projecting white binary speckle pattern,the deformed speckle images were captured and then polar-rectified.Exploiting the improved traditional fast-ZNCC calculation method and zero-mean normalized cross correlation(ZNCC) as the correlation function,the shortcomings of low precision and low efficiency of the conventional stereo vision algorithm for weak texture regional reconstruction were effectively overcame.The developed algorithm was implemented on Myriad2 platforms characterized by heterogeneous multi-core architecture achieving accurate and efficient three dimension reconstruction.The experiments demonstrated that the efficiency and performance in 3D reconstruction were significantly improved without any loss of accuracy assisted by the heterogeneous computational architecture.
引文
[1]Scharstein D,Szeliski R.A taxonomy and evaluation of dense two-frame stereo correspondence algorithms[J].International Journal of Computer Vision,2002,47(1/2/3):7-42.
[2]Cao Xiaoqian.Research on stereo matching algorithms for image pairs[D].Xi’an:University of Chinese Academy of Sciences,2014.[曹晓倩.面向病态场景图像对的立体匹配算法研究[D].西安:中国科学院大学,2014.]
[3]Barrientos B,Cerca M,Garcia-Ma R J.Three-dimensional displacement fields measured in a deforming granularmedia surface by combined fringe projection and speckle photography[J].Journal of Optics A:Pure and Applied Optics,2008,10(10):104027.
[4]Garcia J,Zalevsky Z,Garcia-Martinez P,et al.Three-dimensional mapping and range measurement by means of projected speckle patterns[J].Applied Optics,2008,47(16):3032-3040.
[5]Wu Qing,Liu Senzhen,Huang Xiangsheng,et al.3D motion sensing interaction system based on speckles[J].Journal of Computer-Aided Design&Computer Graphics,2016,28(7):1105-1114.[吴清,刘森镇,黄向生,等.基于散斑的3维体感交互系统[J].计算机辅助设计与图形学学报,2016,28(7):1105-1114.]
[6]Han Lei,Xu Bo,Huang Xiangsheng,et al.Speckle projection systems based on GPU[J].Computer Science,2015,42(8):294-399.[韩磊,徐波,黄向生,等.基于GPU的散斑3维重建系统[J].计算机科学,2015,42(8):294-399.]
[7]Dai Xianqiang,Gai Shaoyan,Da Feipeng.Three-dimensional shape measurement method based on speckle pattern with prior phase information[J].Chinese Journal of Scientific Instrument,2014,35(12):2792-2800.[戴鲜强,盖绍彦,达飞鹏.基于相位信息辅助的散斑匹配3维测量方法[J].仪器仪表学报,2014,35(12):2792-2800.]
[8]Schaffer M,Grosse M,Kowarschik R.High-speed pattern projection for three-dimensional shape measurement using laser speckles[J].Applied Optics,2010,49(18):3622-3629.
[9]Wang G J,Yin X W,Pei X K,et al.Depth estimation for speckle projection system using progressive reliable points growing matching[J].Applied Optics,2013,52(3):516-524.
[10]Chen Zhong,Chen Jiaodou.Full-field vibration measurement based on binocular stereo vision and digital speckle image correlation[J].Journal of Vibration and Shock,2015,34(13):121-126.[陈忠,陈教豆.基于双目立体视觉与数字散斑图像相关的全场振动测量[J].振动与冲击,2015,34(13):121-126.]
[11]Shi Chenbo.Research on fast image registration and high accuracy sub-pixel stereo matching[D].Beijing:Tsinghua University,2011.[施陈博.快速图像配准和高精度立体匹配算法研究[D].北京:清华大学,2011.]
[12]Hirschmuller H,Scharstein D.Evaluation of stereo matching costs on images with radiometric differences[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2009,31(9):1582-1599.
[13]Muhlmann K,Maier D,Hesser J,et al.Calculating dense disparity maps from color stereo images,an efficient implementation[J].International Journal of Computer Vision,2002,47(1):79-88.
[14]Veksler O.Fast variable window for stereo correspondence using integral images[C]//Proceedings of the 2003IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Madison:IEEE,2003:556-561.
[15]Lei Ming,Zhang Guangjun.Image orientation algorithm with subpixel accuracy based on correlative matching method[J].Opto-Electronic Engineering,2008,35(5):108-113.[雷鸣,张广军.基于互相关的图像匹配亚像素定位[J].光电工程,2008,35(5):108-113.]
[16]Xiao Zhitao,Zhang Wenyin,Geng Lei,et al.Accuracy analysis of binocular vision system[J].Opto-Electronic Engineering,2014,41(2):6-11.[肖志涛,张文寅,耿磊,等.双目视觉系统测量精度分析[J].光电工程,2014,41(2):6-11.]
[2]Cao Xiaoqian.Research on stereo matching algorithms for image pairs[D].Xi’an:University of Chinese Academy of Sciences,2014.[曹晓倩.面向病态场景图像对的立体匹配算法研究[D].西安:中国科学院大学,2014.]
[3]Barrientos B,Cerca M,Garcia-Ma R J.Three-dimensional displacement fields measured in a deforming granularmedia surface by combined fringe projection and speckle photography[J].Journal of Optics A:Pure and Applied Optics,2008,10(10):104027.
[4]Garcia J,Zalevsky Z,Garcia-Martinez P,et al.Three-dimensional mapping and range measurement by means of projected speckle patterns[J].Applied Optics,2008,47(16):3032-3040.
[5]Wu Qing,Liu Senzhen,Huang Xiangsheng,et al.3D motion sensing interaction system based on speckles[J].Journal of Computer-Aided Design&Computer Graphics,2016,28(7):1105-1114.[吴清,刘森镇,黄向生,等.基于散斑的3维体感交互系统[J].计算机辅助设计与图形学学报,2016,28(7):1105-1114.]
[6]Han Lei,Xu Bo,Huang Xiangsheng,et al.Speckle projection systems based on GPU[J].Computer Science,2015,42(8):294-399.[韩磊,徐波,黄向生,等.基于GPU的散斑3维重建系统[J].计算机科学,2015,42(8):294-399.]
[7]Dai Xianqiang,Gai Shaoyan,Da Feipeng.Three-dimensional shape measurement method based on speckle pattern with prior phase information[J].Chinese Journal of Scientific Instrument,2014,35(12):2792-2800.[戴鲜强,盖绍彦,达飞鹏.基于相位信息辅助的散斑匹配3维测量方法[J].仪器仪表学报,2014,35(12):2792-2800.]
[8]Schaffer M,Grosse M,Kowarschik R.High-speed pattern projection for three-dimensional shape measurement using laser speckles[J].Applied Optics,2010,49(18):3622-3629.
[9]Wang G J,Yin X W,Pei X K,et al.Depth estimation for speckle projection system using progressive reliable points growing matching[J].Applied Optics,2013,52(3):516-524.
[10]Chen Zhong,Chen Jiaodou.Full-field vibration measurement based on binocular stereo vision and digital speckle image correlation[J].Journal of Vibration and Shock,2015,34(13):121-126.[陈忠,陈教豆.基于双目立体视觉与数字散斑图像相关的全场振动测量[J].振动与冲击,2015,34(13):121-126.]
[11]Shi Chenbo.Research on fast image registration and high accuracy sub-pixel stereo matching[D].Beijing:Tsinghua University,2011.[施陈博.快速图像配准和高精度立体匹配算法研究[D].北京:清华大学,2011.]
[12]Hirschmuller H,Scharstein D.Evaluation of stereo matching costs on images with radiometric differences[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2009,31(9):1582-1599.
[13]Muhlmann K,Maier D,Hesser J,et al.Calculating dense disparity maps from color stereo images,an efficient implementation[J].International Journal of Computer Vision,2002,47(1):79-88.
[14]Veksler O.Fast variable window for stereo correspondence using integral images[C]//Proceedings of the 2003IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Madison:IEEE,2003:556-561.
[15]Lei Ming,Zhang Guangjun.Image orientation algorithm with subpixel accuracy based on correlative matching method[J].Opto-Electronic Engineering,2008,35(5):108-113.[雷鸣,张广军.基于互相关的图像匹配亚像素定位[J].光电工程,2008,35(5):108-113.]
[16]Xiao Zhitao,Zhang Wenyin,Geng Lei,et al.Accuracy analysis of binocular vision system[J].Opto-Electronic Engineering,2014,41(2):6-11.[肖志涛,张文寅,耿磊,等.双目视觉系统测量精度分析[J].光电工程,2014,41(2):6-11.]