光场相机视觉测量误差分析  

Error Analysis of Vision Measurement Based on the Light Field Camera

作　　者：汪义志 张旭东[1] 熊伟 邓武 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009

出　　处：《光子学报》2017年第11期113-124,共12页Acta Photonica Sinica

基　　金：国家自然科学基金(No.61403116);中国博士后基金项目(No.2014M560507)资助~~

摘　　要：光场相机通过一次曝光可以获取空间目标的位置和方向信息,具有重聚焦和多视角的特性,利用光场的这些特性可以进行视觉测量.本文对光场极平面图像视觉测量、重聚焦视觉测量、双目视觉以及多目视觉测量方法的测量原理和误差影响因素进行了理论分析,并通过实验验证了光场视觉测量误差跟不同视角基线长度,主透镜焦距大小,目标离相机的实际距离等结构参量的关系;理论分析和实验结果表明,由于相机基线较短,远距离测量误差较大,近距离测量具有较高的精度;在光场微透镜阵列大小有限条件下,采用多个视角组合的测量方法具有更高的测量精度.Light field camera can obtain position information and direction information of space target through one exposure. It has the characteristics of refocus and multi-view, which can be used for vision measurement. In this paper, the measurement principle and error influence factors of epipolar plane image （EPI）, refocus-based, binocular and trinocular measurement methods were analyzed. The relationship between the measurement errors and the structural parameters, including the baseline length of different views, the focal length of the main lens and the actual distance of the target from the camera, were verified by experiments. Theoretical analysis and experiment results show that, due to the short baseline of the camera, the distance measurement errors are large, and short distance is of high precision; under the condition of fixed size of microlens array, the multi-view combination measurement method has higher accuracy.

关 键 词：光场相机 多视角图像 焦堆栈图像 视觉测量 结构参量 误差分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]