结构光
校准
稳健性(进化)
计算机科学
翻译(生物学)
平面的
投影(关系代数)
计算机视觉
人工智能
数学
算法
计算机图形学(图像)
信使核糖核酸
统计
基因
生物化学
化学
作者
Tianyu Li,Fajie Duan,Xiao Fu,Changwen Liu,Chunjiang Liang,Anqi Chen,Xin Li
出处
期刊:Measurement
[Elsevier]
日期:2023-05-04
卷期号:216: 112930-112930
被引量:8
标识
DOI:10.1016/j.measurement.2023.112930
摘要
Three-dimensional (3D) reconstruction techniques have been widely employed in many fields. But precise endoscopic 3D measurement inside of objects, especially in tight and restricted spaces, is still a challenge since there is no easy access for measurement tools. Endoscope with 2D-view function has been widely used. However, the precise measurement technology of the inner surface of small hollow objects has not been widely researched. The components of the existing endoscope three-dimensional measuring probe are complex, expensive and not performed well in accuracy. Thus, a novel method and a miniaturized endoscopic 3D sensor is introduced. To be suit in tight and restricted space, the sensor probe prototype has a maximum 6 mm in thickness. The miniatured sensor prototype is mainly based on improved line-structured light, which performs better in robustness and measurement accuracy than coded projection. Especially, the components are inexpensive and easily available, which reduces the manufacturing cost thus making it easily to popularize in industry. Correspondingly, to ensure the sensor measurement accuracy, we also proposed a calibration method for a miniature line-structured light system based on specially designed planar targets. It should be noted that the 3D extension of the line-structured light system relies on the movement of the translation stage. Taking full account of the characteristics of the sensor prototype, a high-precision direction calibration method of the translation stage based on joint multi-calibration position noise estimation is proposed. Finally, the accuracy experiment is carried out on the gauge block to verify that the measurement method proposed can achieve high-precision 3D reconstruction in a restricted space, and that the 3D reconstruction error is less than 10µm.
科研通智能强力驱动
Strongly Powered by AbleSci AI