用户登录
期刊信息
  • 主管单位:
  • 上海市教育委员会
  • 主办单位:
  • 上海理工大学
  • 主  编:
  • 庄松林
  • 地  址:
  • 上海市军工路516号
  • 邮政编码:
  • 200093
  • 联系电话:
  • 021-55277251
  • 电子邮件:
  • xbzrb@usst.edu.cn
  • 国际标准刊号:
  • 1007-6735
  • 国内统一刊号:
  • 31-1739/T
  • 邮发代号:
  • 4-401
  • 单  价:
  • 15.00
  • 定  价:
  • 90.00
钱天磊,周骛,马原驰,刘慧芳,蔡小舒.基于虚拟深度的聚焦型显微光场相机深度测量标定[J].上海理工大学学报,2020,42(2):131-137.
基于虚拟深度的聚焦型显微光场相机深度测量标定
Calibration for depth estimation by a focused micro light-field camera based on virtual depth
投稿时间:2019-02-18  
DOI:10.13255/j.cnki.jusst.20190218003
中文关键词:  相机标定  聚焦型显微光场相机  深度测量  多目立体视觉  图像匹配
英文关键词:camera calibration  focused micro light-field camera  depth estimation  multi-view stereo vision  image matching
基金项目:国家自然科学基金资助项目(51576130,51327803);上海市教委晨光计划(14CG46)
作者单位E-mail
钱天磊 上海理工大学 能源与动力工程学院, 上海 200093  
周骛 上海理工大学 能源与动力工程学院, 上海 200093 usst_wzhou@163.com 
马原驰 上海理工大学 能源与动力工程学院, 上海 200093  
刘慧芳 上海理工大学 能源与动力工程学院, 上海 200093  
蔡小舒 上海理工大学 能源与动力工程学院, 上海 200093  
摘要点击次数: 70
全文下载次数: 93
中文摘要:
      针对聚焦型显微光场相机在内部光学参数未知的情况下,进行了基于虚拟深度的深度测量标定。首先基于高斯光学建立光场成像模型,推导出虚拟深度与实际深度间的函数关系。然后选择单一角点的标定板,在不同深度位置进行拍摄;该角点在多个宏像素中重复成像,相邻重复像点的间距随深度位置改变而变化;利用图像匹配的方法计算相邻重复像点的距离和虚拟深度值,并与实际深度一一对应进行曲线拟合。根据拟合结果,分析了不同深度位置下,该光场成像系统的深度测量分辨率。最后,通过拍摄已知倾角的倾斜棋盘格标定板,进行深度测量并分析测量误差,在主镜头工作距离靠近镜头方向2 mm(10倍景深)范围内,测量误差小于5.35%。
英文摘要:
      Without knowing the intrinsic light-field parameters, a calibration process was conducted for depth estimation by a focused micro light-field camera based on virtual depth. In order to derive the functional relation between virtual depth and real depth, a light-field imaging model was established according to the Gaussian optics. A single corner calibration board was imaged at various depth distances. The corner point was multi-imaged in several macro-pixels, and the distance between adjacent redundant image points varied according to the depth value. By using an image matching algorithm, the distance between adjacent redundant image points as well as the virtual depth was calculated. After that, the curve for the relation between the virtual depth and corresponding real depth was fitted. Based on this fitted function, the resolution of depth estimation at different depths was analyzed. By capturing a slanted chessboard target, the depth estimation error was obtained. When the working distance toward the lens is within the range of 2 mm (10 times DOF), the error of depth estimation is lower than 5.35%.
HTML   查看全文  查看/发表评论  下载PDF阅读器