用于聚焦型光场相机标定的棋盘角点检测算法

Checkerboard Corner Detection Algorithm for Calibration of Focused Plenoptic Camera

精确标定能够发挥出聚焦型光场相机在场景重建和非接触测量等方面的作用。而标定精度提升的关键之一在于精确的特征提取算法。为了提升特征检测的精度和效率,提出一种基于原始图的棋盘格角点检测算法。利用稳健的角点检测算子对原始图角点进行检测,并利用二维角点与三维光场圆域特征的对应关系进行角点筛选。然后,利用图像一致性对角点进行亚像素优化。进行了仿真角点检测实验和仿真标定实验,并基于R29聚焦型光场相机得到的重建角点进行了距离测量实验。实验结果表明,所提角点检测算法的精度高于现有算法,并证明了基于所提角点检测方法的标定算法能够得到更加精准的结果。

Accurate calibration can develop the role of a focused plenoptic camera in the fields like scene reconstruction and non-contact measurement.One of the keys to improve the calibration precision is the accurate feature extraction algorithm.In order to improve the accuracy and efficiency of feature detection,we present a checkerboard corner detection algorithm based on the raw images.First,a robust corner detection operator is used to detect the checkerboard corners in the raw images,and the corresponding relationship between the 2D corners and the 3D plenoptic disc features is used to screen the detected results.Then,the sub-pixel optimization is carried out using the image consistency.The simulated corner detection and calibration experiments are carried out,and the distance measurement experiment is also carried out based on the reconstructed corners obtained by the R29 focused plenoptic camera.The experimental results show that the accuracy of the proposed corner detection algorithm is higher than those of the existing algorithms,and the calibration algorithm based on the proposed corner detection algorithm can achieve more accurate results.