Research on vision-based measurement of medium-module spur cylindrical gears

Abstract

Gear accuracy directly affects gear quality and the performance of mechanical transmission. To address the inefficiency of contact measurement for medium-module spur gears and the high cost of precision measuring instruments, this paper proposes a vision-based non-contact measurement method for medium-module spur gears. The method establishes an optical imaging and vision measurement framework tailored to medium-module spur gears, and applies image preprocessing to enhance edge features. Subsequently, gear contour information is extracted using the least-squares method, and precise subpixel localization and deviation measurement of tooth profiles are achieved by combining the involute gear geometry with an improved subpixel edge detection algorithm based on the Bertrand gray surface model. Experiments were conducted on four spur gears with a module of 5 mm. The results show that the calibrated pixel scale of the system is 36.1869 μm/pixel, and the maximum measurement error for the addendum circle radius is 0.342 mm. Furthermore, the tooth profile deviation curves obtained by the vision system closely matched those from a gear measuring center, and the measured deviations exhibit good consistency with the reference measurements. These findings validate the practicality and reliability of the proposed method for geometric parameter extraction and individual deviation measurement of medium-module spur gears.