摘要
基于光场成像技术的场景三维深度重建,需标定光场相机的几何参数。本文提出一种基于原始光场图像的聚焦光场相机的标定方法。拍摄标定板不同旋转角度的原始光场图像;根据图像上像点与虚拟像点(像点关于微透镜的共轭点)的共轭关系,计算得到虚拟像点的坐标;根据标定板上角点与虚拟像点的共轭关系,建立聚焦光场相机标定模型;利用Levenberg-Marquardt算法求解标定模型,进行标定实验;比较所提方法与基于全聚焦图像的标定方法的标定参数。结果表明,原始光场图像与全聚焦图像对应的虚拟像点(角点关于主透镜的共轭点)之间的误差小于21pixel,角点的标定误差小于3%,基于原始光场图像和全聚焦图像方法获得的结构参数和外部参数具有较好的一致性,表明基于原始光场图像的标定方法的可行性。
In the three-dimensional depth reconstruction of the scene based on light field photography, it is necessary to calibrate the geometric parameters of light field cameras. In this paper, a calibration method of focused light field cameras is proposed based on Jiangsu raw light field images. The raw light field images of a calibration board with different orientations are captured. According to the conjugation relationship between image points and virtual image points （conjugation points of image points for the microlens）, the coordinates of the virtual image points are calculated. The calibration model of focused light field cameras is established according to the conjugation relationship between the corner points on the calibration board and the virtual image points. The model is then solved by Levenberg-Marquardt algorithm. Calibration experiments are carried out. The accuracy of the proposed method is compared to that of the calibration method based on the total focused images. Experimental results show that the error between the virtual image points obtained from raw light field images and those （conjugation points of corner points for mainlens） from total focused images is less than 21 pixels. The relative calibration errors of the corner points are less than 3 %. The calibrated configuration parameters are in good consistence with those from total focused images calibrating the focused light field cameras. and external parameters from raw light field images The proposed method is proved to be effective calibrating the focused light field cameras.
出处
《光学学报》
EI
CAS
CSCD
北大核心
2017年第5期176-186,共11页
Acta Optica Sinica
基金
国家自然科学基金(51676044
51506030
51327803)
江苏省自然科学基金杰出青年基金(JL150501)
关键词
机器视觉
聚焦光场相机
全聚焦图像
虚拟像点
标定方法
machine vision
focused light field camera
total focused image
virtual image point
calibration method
