In this paper, a method is proposed that improves the focal length estimation in an augmented reality surgical guidance system for procedures involving a stereoscopic surgical microscope. Firstly, we show the sensitivity of a photogrammetric calibration method towards the detection of 2D markers in the projected calibration image and the markers' positional accuracy on the calibration object, both of them affecting the accuracy of the focal length estimate. Secondly, we propose a hybrid method using a photogrammetric method for pre-calibration and a number of self-calibration based methods to further optimise the focal length estimate., The results of the reported experiment, evaluating the proposed method, show a significant improvement in the calibration error of around 0.2 pixels as compared to calibrating each camera separately using a photogrammetric method only.
展开▼