基于双目图像融合的孔轴配合间隙监视

Inspection of Hole-shaft Assembly Clearance Based on Binocular Image Fusion

孔轴配合常见于工业制造业的生产活动中,实现装配自动化、提升装配效率需要以视觉监视技术作为支撑。该文提出一种基于双目图像融合的孔轴配合间隙监视技术。通过融合双目视觉系统在两个视角下拍摄的孔轴配合图像,可以有效解决单视角下的图像遮挡和视场受限问题,从而有效消除视场盲区。虚拟交线指的是轴件与圆孔支撑平面之间的交线,该交线在图像上并不可见。在获得双目融合图像后,通过构造并提取图像中二次曲线相切不变量,结合Nelder-Mead单纯形搜索法对虚拟交线几何参数进行迭代优化,迭代优化后的虚拟交线与圆孔边缘轮廓间的相对位置关系反映了该平面上完整的孔轴配合间隙。实验结果表明,提出的方法可以对配合间隙为0.5 mm及以下的孔轴配合过程进行监视。

Hole-shaft assembly is common in the manufacturing industry.It takes visual inspection as a support to implement assembly automation and improve efficiency of assembly.We propose a vision-based method involving the fusion of binocular images to inspect clearance between the hole and the shaft.The problem of image occlusion and restricted field of view would be solved by fusing two images captured under dual perspectives,and the blind zone could also be reduced.The virtual intersection between the shaft and the supporting plane of the hole is invisible in images.After creating and extracting the invariance of tangence in the fused image,the geometric parameters of the virtual intersection could be optimized through the Nelder-Mead method.And then,the virtual intersection and the contour of the hole could present complete information about the clearance between the hole and the shaft in the supporting plane.The experiment shows that the proposed method can inspect the hole-shaft assembly process with the clearance 0.5 mm or below.