计算机应用 ›› 2013, Vol. 33 ›› Issue (08): 2300-2305.

• 多媒体处理技术 • 上一篇    下一篇

改进基本矩阵计算和优化的多摄像机并行标定算法

李斌1,2,谭光华1,2,高春鸣1,2   

  1. 1. 湖南大学 数字媒体研究所,长沙 410082
    2. 湖南大学 信息科学与工程学院,长沙 410082
  • 收稿日期:2013-02-23 修回日期:2013-04-04 出版日期:2013-08-01 发布日期:2013-09-11
  • 通讯作者: 谭光华
  • 作者简介:李斌(1987-),男,湖南衡阳人,硕士研究生,主要研究方向:摄像机定标、机器视觉;
    谭光华(1982-),男,湖南邵阳人,讲师,博士,主要研究方向:数字几何处理、逆向工程;
    高春鸣(1957-),男,山东胶州人,教授,博士,主要研究方向:数字媒体、服务计算。
  • 基金资助:
    广东省教育部产学研结合项目

Improved foundamental matrix computation and optimization method in parallel calibration of multi-camera system

LI Bin1,2,3,TAN Guanghua1,2,3,GAO Chunming1,2,3   

  1. 1. Institute of Digital Media,Hunan University, Changsha Hunan 410082,China
    2. College of Information Science and Engineering, Hunan University,Changsha Hunan 410082,China
    3. Institute of Digital Media,Hunan University, Changsha Hunan 410082,China
  • Received:2013-02-23 Revised:2013-04-04 Online:2013-09-11 Published:2013-08-01
  • Contact: TAN Guanghua

摘要: 多摄像机系统具有摄像机数目多、空间位置分布复杂特点,导致多摄像机标定效率低。基本矩阵计算和非线性优化是摄像机标定算法的关键步骤。针对标定物空间位置相互独立性,改进随机抽样一致性(RANSAC)的基本矩阵计算和简化非线性优化的增量方程,提出多摄像机系统的并行标定算法。该算法挖掘多摄像机标定过程的内在并行化,从而提高了标定的时间效率。相比于传统的多摄像机标定算法,并行算法的时间复杂度从O(n3)降为O(n)。实验结果表明:使用多摄像机系统并行标定算法在不损失精度的同时能够减少标定时间,实现多摄像机系统的快速标定。

关键词: 摄像机定标, 随机抽样一致性, 参数独立性, 捆绑调整, 并行计算

Abstract: Multi-camera system has a number of cameras and complex space distribution, hence the calibration in multi-camera system is inefficient. The calculation of fundamental matrix and nonlinear optimization are the key steps in camera calibration algorithm. In the light of independent marked points in multi-camera system calibration, fundamental matrix based on RANdom SAmple Consensus (RANSAC) and increment equation in nonlinear optimization were improved, and a parallel calibration algorithm for multi-camera system was proposed. Based on the analysis of the parallel process of calibration, this algorithm improved the efficiency of the calibration time. Compared with the traditional camera calibration algorithm, it makes the time complexity of calibration reduce from O(n3) to O(n). The experiments show that the parallel algorithm reduces the time obviously without any loss in precision, thus realizing fast multi-camera system calibration.

Key words: camera calibration, RANdom SAmple Consensus (RANSAC), parameters independence, bundle adjustment, parallel computation

中图分类号: