Journal of Frontiers of Computer Science and Technology ›› 2020, Vol. 14 ›› Issue (7): 1104-1113.DOI: 10.3778/j.issn.1673-9418.2002030

Previous Articles     Next Articles

Research and Development of Binocular Stereo Matching Algorithm

ZHAO Chenyuan, LI Wenxin, ZHANG Qingxi   

  1. Lanzhou Institute of Physics,Lanzhou 730000, China
  • Online:2020-07-01 Published:2020-08-12

双目视觉的立体匹配算法研究进展

赵晨园李文新张庆熙   

  1. 兰州空间技术物理研究所,兰州 730000

Abstract:

Binocular stereo vision has the advantages of low cost and wide applicability. It is extensively used in object recognition, object detection and has become the research focuses in the field of computer vision. Stereo matching is the critical process in the binocular stereo vision system, which plays an important role in the realization of three-dimensional reconstruction. The practical application of matching cost computation, cost aggregation, disparity computation/optimization and disparity refinement is described and the future trend of stereo matching is summarized and prospected based on the analysis of local, global, semi global and depth learning based stereo matching algorithms. Stereo matching will focus on solving the problem of mismatches in non-occluded and textureless regions. Deep learning algorithm for stereo matching has the advantages of high precision, and which is the future trend. Stereo matching algorithm will advance toward high precision, and real-time. Ideas and references are provided for the develop-ment of stereo matching of binocular vision technology.

Key words: binocular stereo vision, stereo matching, local stereo matching, global stereo matching, semi-global stereo matching, deep learning

摘要:

双目立体视觉技术具有成本低、适用性广的优点,在物体识别、目标检测等方面应用广泛,成为计算机视觉领域的研究热点。立体匹配是双目立体视觉技术中的核心算法,对实现物体三维重建具有重要作用。在分析局部、全局、半全局和基于深度学习的立体匹配算法研究现状的基础上,阐述了匹配代价、代价聚合、视差计算/优化、视差校正在立体匹配算法中的实际应用,总结及展望了立体匹配算法的发展趋势。立体匹配算法将着重解决无、弱纹理区域误匹配问题,深度学习算法用于立体匹配具有精度高的优点,是未来发展趋势,匹配算法将向高精度、实时性方向发展,为双目视觉技术中立体匹配算法的发展提供思路与参考。

关键词: 双目立体视觉, 立体匹配, 局部立体匹配, 全局立体匹配, 半全局立体匹配, 深度学习