Distortion estimation to improve tsai calibration for cameras with large optical distortion

Tsai's calibration has several advantages over the current state-of-the-art, Zhang calibration, when it comes to field-work and calibration systems requiring little user interaction. Unfortunately Tsai's calibration performs less well against cameras with large optical distortion, such as those used in conjunction with Simultaneous Localization and Mapping (SLAM) systems. The goal of this work is to solve this problem by estimating distortion from a Tsai calibration object prior to calibration. The proposed approach implicitly preserves the pin-hole model within a homography representation and determines distortion based on the linearly of the system of equations linking calibration object feature points to image feature points. The experimental results presented in this work demonstrate that the proposed approach leads to more accurate calibrated models and inter-frame models than current existing approaches with respect to Tsai calibration.