Kontinuierliche Selbstkalibrierung von Stereokameras

This article describes an algorithm for continuous self-calibration of stereo cameras and highlights results obtained in [1]. The contribution of this work is a detailed analysis of the influence of individual camera parameters on three-dimensional reconstruction and a theoretically consistent presentation of all processing steps required for stereo self-calibration. Different geometric constraint equations – such as the epipolar constraint between corresponding points in two camera images, the trilinear constraints between image triplets and bundle adjustment for long image sequences – are formulated in a common Gauss-Helmert type model, which allows calibration with high accuracy. Estimates of the camera parameters are obtained using a robust, recursive filter algorithm for Gauss-Helmert models. This algorithm is adapted from an iterative extended Kalman filter. It is shown that a combination of different constraint equations increases the reliability of parameter estimation. The performance of the continuous self-calibration is demonstrated on real imagery of an active stereo camera system.