计算机科学
校准
极线几何
投影(关系代数)
结构光三维扫描仪
轮廓仪
失真(音乐)
人工智能
计算机视觉
棋盘
算法
数学
工程类
带宽(计算)
放大器
扫描仪
几何学
图像(数学)
统计
机械工程
表面光洁度
计算机网络
作者
Jian Wang,Zonghua Zhang,Wenlong Lu,Xiangqian Jiang
出处
期刊:IEEE-ASME Transactions on Mechatronics
[Institute of Electrical and Electronics Engineers]
日期:2022-01-17
卷期号:27 (5): 4199-4204
被引量:34
标识
DOI:10.1109/tmech.2021.3136617
摘要
Calibration is a fundamental step in fringe projection profilometry development. Among various methods, hybrid calibration is a promising solution given its high flexibility and sensing speed. However, existing approaches suffer the problem of limited accuracy and convenience. For example, customized high-accuracy artifacts with complex operations are usually required. This article reports a hybrid method with improved accuracy and operability, based on a unique back-projection and nonlinear epipolar sampling analysis. This method is performed in three steps: stereo calibration, residual distortion analysis, and phase 3-D mapping tabulation; the last two are pure computations without manual operations. The method is straightforward and cost-effective since only a checkerboard-based stereo calibrating operation is required. It is flexible without checkerboard poses to be known and also highly accurate because systematic distortions are almost compensated. The eventually obtained phase 3-D mapping table is computationally efficient for high-speed sensing. Our experiments confirm that a relative measurement error of 0.01% relative to the field-of-view has been finally achieved.
科研通智能强力驱动
Strongly Powered by AbleSci AI