期刊文献+

基于卷积神经网络的半全局立体匹配

下载PDF
导出
摘要 卷积神经网络在训练时,输入的图像对可能存在尺寸不一致,以及半全局算法的路径代价聚合不能完全利用待匹配点周围的像素,本文提出一种结合空间金字塔的网络结构训练图像对,计算初始匹配代价,在代价聚合时采用一种结合树形结构的半全局算法。改进后的算法通过实验证明,误匹配率有了明显的降低,获取的视差图精度更高。
作者 谭浩 车志龙 苏翠丽 TAN Hao;CHE Zhi-long;SU Cui-li
出处 《信息技术与信息化》 2019年第11期137-138,共2页 Information Technology and Informatization
  • 相关文献

参考文献3

二级参考文献34

  • 1Hartley R,Zisserman A.计算机视觉中的多视图几何[M].韦穗,杨尚骏,译.合肥:安徽大学出版社,2002.
  • 2Geiger D, Ladendorf B, Yuille A. Occlusions and binocular stereo [J]. International Journal of Computer Vision, 1995, 14(3): 211-226.
  • 3Gong M, Yang Y H. Near real-time reliable stereo matching using programmable graphics hardware [J]. IEEE Computer Society Conference on ComputerVision and Pattern Recognition, 2005, (1): 924-932.
  • 4Veksler O. Stereo correspondence by dynamic programming on a tree [J]. IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2005, (2): 384-390.
  • 5Birchfield S. Depth discontinuities by pixel-to-pixel stereo [J]. Intemational Journal of Computer Vision, 1999, 35(3): 269-293.
  • 6Daniel S, Richard S. A taxonomy and evaluation of dense two-frame stereo correspondence algorithms [J]. International Journal of Computer Vision, 2002, 47(1): 7-42.
  • 7Boykov Y, Veksler O, Zabih R. Fast approximate energy minimization via graph cuts [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2001, 23(11): 1222-1239.
  • 8Salmen J, Schlipsing M, Edelbrunner J. Real-time stereo vision: making more out of dynamic programming [C]//. Computer Analysisi of Images and Patterns, beRLIN: 2009: 1096-1103.
  • 9Humenberger M, Zinner C, Weber M. A fast stereo matching algorithm suitable for embedded real-time systems. Computer Vision Image Understanding, 2010, 114(11): 1180-1202.
  • 10Gargallo P,Sturm P F. Bayesian 3D modeling from images u-sing multiple depth maps [ C ]//Proc of CVPR. [ s. 1. ] : [ s. n. ] ,2005:885-891.

共引文献6

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部