【作者】 陈力；

【导师】 朱建华；

【作者基本信息】 四川大学 ， 光学， 2006， 硕士

【摘要】 在众多的三维面形测量方法中，激光三角测量法作为一种非接触式测量方法，有着光学结构简单、速度快、可在线测量、成本低、稳定性好等优点，目前已经在三维面形测量、快速成型、逆向工程上得到了广泛的应用。近年来，随着光学和微电子技术的发展，便携式集成三维扫描系统得到了快速发展，它可以方便地从多角度多方位获取物体的三维信息，已经在太空探测和机器人视觉方面得到了较好的应用。但是当激光视觉测量系统应用于一些尺寸较大、面形较复杂的物体的三维检测时，由于扫描系统景深的限制，测量精度常常不高，必须进行多路径编辑，并对每一次扫描的数据进行叠加，这将增加激光扫描的时间以及增大三维重建时的复杂程度，降低工作效率，因此增大扫描系统的景深对于提高测量精度、降低三维重建时拼接过程的复杂度，以及提高效率等有着重要的意义。 本文中，我们基于Scheimpflug条件对CCD平面进行偏转，成功地将激光扫描系统的景深由30mm扩大到100mm以上；通过建立CCD偏转后的摄像机模型，推导了CCD偏转后的物像关系方程，对扩大景深后的激光扫描系统进行了标定；同时针对景深扩大后带来的测量精度显著降低的问题提出了一种将系统全景深范围进行分段校准的新方法，使用标定得出的摄像机参数进行三维测量时的精度可以达到0.06mm。在以上研究工作的基础上，成功开发出了一种景深范围大、测量精度高的便携式三维激光扫描系统。 为将所开发的大景深便携式三维激光扫描仪用于机器人视觉，我们提出了一种将一个已知半径的标准球作为工具的机器人“手-眼”标定的新方法，完成了机器人的“手-眼”标定实验和三维重建实验，实验证明机器人能够与所开发的便携式三维激光扫描仪一起很好的完成高精度、高稳定性的三维扫描任务，其三维重建误差小于0.2mm。更多还原

【Abstract】 Laser-scanning technology based on the triangulation method has many advantages such as non-contact operation, fast acquisition speed, simple optical arrangement, low cost and good stability, etc., now it has been widely applied in the measurement of surfaces and molds, as well as in prosthesis design, rapid prototyping, and reverse engineering, etc. In recent years, with the development and progress in optical and microelectronic technologies, compact and portable 3D acquisition systems are integrated for acquiring the 3-D information of the object’s surface from multiple angles and directions, and have been used for the space applications and robot vision. However, when it is applied to the measurement of object with large size and complicated profile, the limitation of small depth-of-field of conventional 3D acquisition system will usually lead to the low measurement precision. One has to edit multiple scanning paths and overlap corresponding scanning data to improve the measurement precision, but this process is time-consuming and inefficient.In this paper, in order to extend the depth -of field of 3D scanner, we tilt the Charge Coupled Device (CCD) plane according to the Scheimpflug condition, then construct the imaging model of the camera and theoretically derive the object-image equations of tilted camera system. We extend the depth -of field of 3D scanner from 30mm to 100mm by camera calibration experiments. Meanwhile, a segmental calibration method is proposed to effectively improve the measurement precision of the scanning system. Finally, a large depth-of-field portable three-dimensional laser scanner with measurement precision of 0.06 mm is developed successfully.In order to apply the large depth-of-field portable three-dimensional laser scanner to robot vision, a new eye-in-hand calibration method using a criterion ball with a known radius is proposed. Using this method, we successfully accomplished the object’s 3D reconstruction experiment. It showed that 3D scanning results with high precision and good stability can be readily obtained using robot and the large depth-of-field portable three-dimensional laser scanner developed by ourselves, the maximum error is less than 0.2mm.更多还原