Hough transform guided subpixel edge regression for high precision relocalization of nut geometric parameters

Abstract

This study addresses the challenge of multiple geometric dimension measurement of nut based on a dual-position opto-electronic imaging platform, and the hardware's core function lies in the rapid acquisition of image information, while the software primarily provides high-precision image measurement algorithms. This study proposes a series of algorithms including preprocessing for noise elimination and illumination improvement, integer edge point extraction, sub-pixel edge point extraction, and Hough Transform-guided sub-pixel edge point-based line segment and ellipse relocation—a decisive algorithm for high-precision measurement—along with a geometric parameter calculation model. These innovations collectively enable the computation of four key metrics: Opposite Side Distance Range (OSDR), Thickness Range (TR), Thread Diameter Range (TDR), and Least Diagonal Distance (LDD). The developed optoelectronic instrument achieved an average measurement accuracy of 0.05 mm due to a 97 % confidence level from M5 to M16 nut specifications. Meanwhile, the instrument's two primary evaluation metrics—Standard Deviation of Machine Measurement (SDMM) and Standard Deviation of Measurement Error (SDME)—are 0.007 mm and 0.020 mm respectively, reflecting its stability and reliability. Extensive experimental testing and statistical analysis across various specifications and measurement standards have demonstrated that the geometric parameter relocation method based on sub-pixel edge point regression is a critical pathway for achieving high-precision measurement.