Measurement of perpendicularity of flexible joint based on machine vision

This paper proposes an accurate machine vision-based method for detecting the end face perpendicularity of flexible joints. Addressing the challenge of angular deviation between the actual reference axis and the vertical axis during inspection, the method extracts the edge contour of the flexible joint's thin rod using adaptive region positioning, edge detection, inlier selection, and line fitting. A virtual central axis is computed for image correction. Subsequently, an adaptive edge region positioning algorithm is used to accurately and efficiently locate the end-face region and the vertical edge region of the outer flexible head. Next, an edge contour processing algorithm is employed to achieve robust edge extraction. Perpendicularity is then calculated based on the offset of virtual feature points along the Y-axis coordinate. Experimental results show a repeatability accuracy of 0.007 mm, with deviations from the traditional method ranging from 0.0007 mm to 0.0097 mm.