基于线结构光的铰链几何参数高精度测量系统

发布时间：2019-06-20 18:15

【摘要】：通过结构光实现非接触式测量的机器视觉技术已经得到广泛应用。本文设计的高精度测量系统旨在获得某型号冰箱铰链上多种特征的形状大小、位置关系以及角度关系,在实验室原有技术积累的基础上做了如下工作:系统通过沿固定方向扫描方式获取待测工件的三维形貌信息,在扫描过程中通过自主设计的基于旋转编码器的闭环反馈触发电路实现相机触发拍照。该触发系统打破了被测物、相机必须在固定位置获取扫描绝对距离这一限制,在扫描过程中被测物持续运动直至扫描结束,提高了扫描速率并降低了抖动。经过摄像机标定后,可求得线结构光条上的点在摄像机归一化坐标系中的坐标,结合线结构光平面方程可得到其在摄像机坐标系下的坐标,光平面方程是完成上述转换的关键。此前采用交比不变法,本文通过将标定板平面、光平面、成像投影平面之间的位置关系完成光平面标定,增加了参与平面拟合的数据量,经验证该算法具有更高的精度和鲁棒性。将扫描过程各个坐标系融合到统一坐标系下需求得各局部坐标系的平移向量,平移向量是扫描步进距离和方向余弦的数乘。此前采用虚拟双目立体视觉标定法获取方向余弦,本系统通过连接多个标定位置的内角点形成多条与扫描方向平行的直线,据此求出扫描方向余弦,从而提高了标定精度。在获取待测工件整体三维形貌点云后,利用开源三维点云处理库PCL(point cloud library)得到所需的特征点云(圆柱、平面、圆),通过轴线向量、表面法向量计算各角度关系、位置关系等。最终测量结果表明,闭环反馈触发系统能够在预期位置完成触发,光平面标定和方向余弦标定具有较高精度,PCL能够较好地完成点云分割和特征拟合,整体测量精度有所提高。
[Abstract]:Machine vision technology for contactless measurement through structured light has been widely used. The purpose of the high precision measurement system designed in this paper is to obtain the shape, position and angle relations of various features on the hinge of a certain type of refrigerator. on the basis of the accumulation of the original technology in the laboratory, the system obtains the three-dimensional morphology information of the workpiece to be measured by scanning along the fixed direction, and realizes the camera trigger photography through the closed-loop feedback trigger circuit based on rotating encoder designed by ourselves in the scanning process. The trigger system breaks the limit of the measured object and the camera must obtain the absolute distance of the scan in a fixed position. During the scanning process, the measured object moves continuously until the end of the scan, which improves the scanning rate and reduces the jitter. After camera calibration, the coordinates of the points on the line structured light strip in the camera normalized coordinate system can be obtained, and the coordinates in the camera coordinate system can be obtained by combining the line structured light plane equation. The optical plane equation is the key to complete the above transformation. Before using the cross ratio invariant method, this paper completes the calibration of the optical plane by the position relationship between the calibration board plane, the optical plane and the imaging projection plane, and increases the amount of data involved in the plane fitting. It is proved that the algorithm has higher accuracy and robustness. The translation vector of each local coordinate system needs to be obtained by fusion of each coordinate system into the unified coordinate system. The translation vector is the number multiplication of the scanning step distance and direction cosine. Previously, the virtual binocular stereo vision calibration method was used to obtain the directional cosine. By connecting the inner corners of multiple calibration positions, the system formed several straight lines parallel to the scanning direction, according to which the scanning direction cosine was obtained, thus the calibration accuracy was improved. After obtaining the whole 3D morphology point cloud of the workpiece to be measured, the required feature point cloud (cylinder, plane, circle) is obtained by using the open source three-dimensional point cloud processing library PCL (point cloud library), and the angle relations and position relations are calculated by axis vector and surface normal vector. The final measurement results show that the closed-loop feedback trigger system can complete the trigger in the expected position, the optical plane calibration and directional cosine calibration have high accuracy, PCL can complete the point cloud segmentation and feature fitting, and the overall measurement accuracy has been improved.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】