Curvature-Based Machine Vision Method for Measuring the Dimension of Ball Screws

Abstract

As a linear actuator, accurate dimension measurement is crucial to the transmission reliability and interchangeability of ball screws. However, most of the current approaches are ineligible for rapid ball screw in-situ inspections due to the installation condition requirement of the production line. In this research, a machine vision method is presented to achieve highly accurate measurements of crucial parameters (the center distance and raceway arcs) in ball screws using a curvature edge detection algorithm. To capture images of the immediate area surrounding the area of interest, a telecentric lens is used. Thereafter, the curvature-based edge detection algorithm is employed to extract the contours. The measurement location on the object is automatically chosen by using a shape-matching algorithm. Additionally, random noise is suppressed by using the multiple-measurement averaging technique. Based on the results of the experiments, it is concluded that the center distance and the two raceway arcs computed absolute errors are 0.0019 mm, 0.0055 mm, and 0.0059 mm, respectively.

Keywords

• Ball screw
• curvature
• edge detection
• machine vision
• shape matching