Global Calibration of Double Sensors Vision Measurement System Based on Planar Mirrors

Wanning Zhang, Haishu Tan, Fuqiang Zhou, Yang Liu

Abstract: Limited by the measurement method and the shape of the measured object, the three-dimensional (3D) scanning equipment can only obtain the point cloud within a limited range of the object. Therefore, a measurement system with two vision sensors placed asymmetrically orthogonally is constructed for the 3D defect detection of industrial parts. Using a single planar target directly for global calibration has low calibration accuracy, so the existing methods typically rely on high-precision external auxiliary devices. The goal of this paper is to estimate the relative pose between the two sensors with a large working distance difference and an angle of approximately 90 degrees between the two optical axes. A mathematical model of the feature extraction error of circle array target in various poses is established. The influence of different target poses on the accuracy of extrinsic parameters solution is comprehensively evaluated using error propagation theory. A global calibration method based on planar mirrors is proposed for a double sensors vision measurement system. This method solves the problem that two cameras cannot simultaneously acquire good images of the same target. Moreover, it does not need to use complex targets and high-precision measurement equipment. The calibration process is flexible and simple, and high calibration accuracy can be achieved. The experimental results show that the translation measurement errors are 0.088 mm, 1.102 mm and 0.164 mm respectively, and the rotation angle measurement errors are 0.057 degrees, 0.013 degrees and 0.014 degrees respectively.