基于单目结构光的大物体三维测量关键方法的研究

Research on critical method of large objects three-dimensional measurement based on monocular structured light

  • 摘要: 在单目结构光的三维测量系统中,由于投影仪倾斜投影, 参考平面上的条纹周期展宽, 给测量带来误差,降低了测量精度。同时受大物体自身几何和形貌等因素的影响,以及相交轴测量系统的限制,其单幅面测量范围受限,很难一次测量大物体完整的三维形貌,而且在测量大物体时,摄像机镜头非线性畸变也影响测量精度。根据参考平面上光栅条纹的周期变化规律,提出了一种适用性好、方便快捷的条纹周期校正的理论模型, 在此基础上, 提出了基于条纹周期校正的四步相移法的理论模型,进而提出了基于条纹周期校正的时间相位展开法的理论模型。采用摄像机镜头非线性畸变校正模型,提高测量精度。在被测物表面粘贴标志点,获取其三维坐标,利用SVD分解和L-M优化算法求取转换矩阵,并在设定的全局坐标系下实现三维图像拼接,采用线性加权算法,对重叠区域进行图像融合。实验结果表明,X轴的拼接误差为0.14 mm,Y轴的拼接误差为0.16 mm,Z轴的拼接误差为0.19 mm,其拼接误差均在测量误差允许范围之内。

     

    Abstract: In three-dimensional measurement system based on structured light, the grating fringe cycle is broadened on the reference surface with the reason of the oblique-angle, which brings errors to the measurement and reduces the accuracy. At the same time, because of large objects' geometry and morphology and other factors, as well as the limit of the intersecting axis projection system, the single shape measuring range is limited. It is difficult to measure large objects' complete shape in one time. And when measuring large objects, camera lens' nonlinear distortion also affects the measurement accuracy. According to the periodic pattern of grating stripes on the reference plane, a well-fitted, convenient and quick stripe cycle correction method was proposed. Based on the cycle correction method, a theoretical model of four-step phase shift method was put up, and then a theoretical model of time phase unwrapping based on fringe period correction was proposed. After that, the lens distortion correction model was used to improve the measurement accuracy. With the mark points pasted on the surface of objects, the three-dimensional coordinates of them were got, as well as the transformation matrix by using the singular value decomposition (SVD) and Levenberg-Marquardt (L-M) optimization algorithm, and three-dimensional image mosaic was realized under the global coordinate system. Finally, the linear weighting algorithm was used to realize the image fusion of overlapping areas. Experimental results show that the registration errors in x, y, z axes are 0.14 mm, 0.16 mm, and 0.19 mm respectively, which all meet the requirements.

     

/

返回文章
返回