基于机器视觉的索缆六自由度位移测量方法研究

Displacement is an important parameter to study the dynamic behavior of cable, and machine vision based approaches can provide a convenient， efficient and economical displacement measurement. Nevertheless, it is challenging to employ vision-based system for the cable displacement measurement, due to the unique torsion of cable, and the difficulty to find any two relatively static points on the cable. In this proposal, a two-dimensional coded cylinder-shaped target is proposed to enable the continuous tracking of a monitored point on the cable. An antinoise coding target is developed to improve the quality of captured images. Combined with sub-pixel positioning algorithm, model information of targets, and intelligent algorithms, the way to localize the feature points precisely from an unsharp image of small target is sought. The relationship between the calibration precision and the number of camera self-calibration parameters is studied, based on which a two-step calibration method is proposed to achieve high-precision, flexible calibration of binocular vision system. Optimization methods are discussed based on swarm intelligence algorithms. Then, the three-dimensional spatial coordinates of the feature point is reconstructed. Based on rigid transformation theory and the Euler angles theory, one could transform the three-dimensional displacements of a group of feature points into the 6DOF displacement of the monitored point on the cable, with the assistant of nonlinear optimization method. Prototype will be developed for testing purpose. After all of these efforts, it is expected to greatly facilitate the displacement measurement, and provide accurate data for the cable mechanical behavior study as well as for the motion control.

位移是研究索缆动力学行为的重要参量。机器视觉可提供高效且经济的位移测量方案。索缆扭转导致的遮挡问题以及现场难以安装系统标定所需的标定物等情况，给机器视觉系统的应用提出了挑战。本项目提出借助带编码的筒状标志物，视觉系统可实现对索缆待测点的持续跟踪；研制有一定抗噪能力的编码标志物以提高图像质量；结合亚像素定位算法、标志物模型信息及智能算法，研究小目标模糊图像上特征点的高精度定位方法；探索基于运动柔性体上稀疏特征点的摄像机在线标定方法，采用分步方式来实现高精度、灵活的双摄像机系统标定，并结合群智能算法以优化标定精度；实现特征点三维空间坐标的高精度重构；基于刚体变换及欧拉角理论，建立标志物上特征点的三维位移与待测点六自由度（6DOF）位移之间的转换模型，研究存在噪声时的非线性寻优法，实现6DOF位移测量；研制出完成原理样机。有望简化位移监测实验，提高精度，为索缆力学行为及运动控制研究奠定实验基础。

机器视觉可提供高效且经济的位移测量方案，但索缆扭转导致的遮挡问题以及应用现场难以安装标定相机系统所需的二维标定物等情况，给机器视觉系统在索缆扭转位移测量上的应用提出了挑战。本项目研究了基于机器视觉的扭转位移测量方法的可行性。具体地，本项目系统研究实现了特征点的二维编码的自动编码方法、带编码筒状标志物的自动检测、基于带编码的筒状标志物对旋转运动待测点的持续跟踪，研究了特征点的自动检测方法，实现了基于一维线性标志物的摄像机标定方法，并结合群智能算法、参数更新及归一化方法大幅提升了相机一维标定的精度；实现了特征点三维空间坐标的高精度重构；基于差分演化的方式对特征点进行圆柱拟合，并基于刚体变换及欧拉角分解理论，建立标志物上特征点的三维位移与待测点六自由度位移之间的转换模型，初步获得了平动及扭转位移信息。依托本课题，共培养硕士研究生11名，以第一作者或通信作者身份发表论文8篇，其中SCI一区论文3篇，申请发明专利5项。本研究工作获专家同行认可，获得在国家自然科学基金委的“第一届土木工程青年论坛”上做口头报告的交流机会。