Single-image calibration method for multiple virtual binocular vision system

Abstract

The virtual binocular vision system (VBVS) based on single camera and mirrors has been paid increasing attention, with the advantages of high flexibility and compactness. However, in its calibration process, the conventional methods require several images and obtain different intrinsic parameters, which should be identical between different virtual cameras. To simplify the calibration process and obtain identical intrinsic parameters, a single-image calibration method for the VBVS based on one camera and mirrors is proposed. Through the single image including multiple sub-images of the calibration planes, the intrinsic parameters and distortion coefficients can be obtained. Then, the global calibration of multiple virtual binocular structures can be performed efficiently. To test the performance of the proposed approach, a VBVS composed of four pairs of virtual binocular structures was designed and the calibration experiments were performed using an artificial three-dimensional calibration target. Moreover, an accuracy comparison with the conventional method was carried out. The results exhibited the feasibility of this method and proved its accuracy.