A Robot-Assisted Back-Imaging Measurement System for Transparent Glass

Though the 3-D sensing technique has achieved many successes in various applications, it is still a challenge for commercialized sensors to measure the 3-D shape of a transparent, mirror-like reflection-based glass surface. The reason is that the well-known triangle-based structure lighting method is only valid for parts with diffuse reflection property. To measure the 3-D profile of a glass surface, specially an automotive windshield, a new robot-assisted back-imaging system is introduced in this paper. Using the mirror property of the glass, encoded patterns are projected from an LCD monitor and recorded by two cameras. The distortion of structured light patterns, caused by the curvature of the glass, is used to calculate the 3-D shape. Different from the traditional structured light methods, the back-imaging system can simultaneously calculate both the coordinates and normal of the surface, which is particularly required by the glass industry. Furthermore, this paper also introduces a robot planning/control system that integrates the measurement constraints, vision model of the new back-imaging sensor, and the computer-aided design model together to automatically move the back-imaging sensor for 3-D shape measurement. The proposed methods have been implemented, and the sensor is installed on an ABB robot for measuring an automotive side-door glass. Experimental results are reported and analyzed.