Research on the Calibration Method of Welding Inspection Robot's Binocular Vision

Abstract

In the context of intelligent manufacturing, the automation and intelligent upgrading of welding inspection have become core requirements in industrial fields. Traditional manual inspection faces challenges such as low accuracy, poor efficiency, and difficulties in adapting to hazardous environments. The integration of machine vision and welding robots provides a key solution to these issues, with precise calibration of binocular vision systems serving as the core prerequisite for this integration. This study systematically investigates the calibration of binocular vision systems in welding inspection robots: it completes the resolution of camera internal and external parameters; experimental verification using MATLAB calibration tools demonstrates a reprojected error of 0.06 pixels, with actual measurement errors ≤0.02mm and a standard deviation of 0.0073mm, indicating high calibration accuracy. The research provides a complete and feasible calibration technical solution that enables precise 3D positioning of welds, enhances the automation and intelligence of welding inspection, and can be extended to various industrial inspection scenarios, demonstrating strong engineering application value.