Multi-feature based boresight self-calibration of a terrestrial mobile mapping system

This paper presents a multi-feature based system calibration method for estimating the boresight angles of a land-based mobile mapping system (MMS) comprised of multiple two dimensional (2D) scanners. The method invokes a least-squares adjustment (LSA) to simultaneously estimate several sets of boresight angles for multiple laser scanners incorporated in an MMS as well as the parameters associated with one or more types of geometric features. This is achieved by constraining the groups of feature point clouds captured by multiple runs to fit their corresponding geometric models in such a way that the weighted sum of squares of adjustment residuals is minimized. The method is particularly suitable for in situ calibration because the geometric features involved are commonly occurring structures (e.g. building façades, bridge surfaces, highway signs and hanging power cables) that are usually captured during the actual survey. In addition to using a planar feature model for calibration, a novel and rigorous three-dimensional (3D) catenary curve model is proposed for geometric modelling of hanging cables to augment the calibration. The proposed calibrations were examined with several different combinations of groups of planar and catenary features and the resulting analysis suggests that the in situ calibrations are effective when compared to the manufacturer's dedicated calibration, with the overall point cloud accuracies for plane fitting being 5.5 cm and 5.4 cm for the vertical and horizontal directions, respectively. It has been successfully demonstrated that the proposed method can be used in a scene having no building façades but only some long hanging cables and horizontal ground surfaces. This is particularly useful for rural areas or inter-city/provincial highways where building façades cannot commonly be captured. Parameter correlations in the calibrations were also addressed. It has also been shown that using catenary features in addition to planar features for the calibration can help de-correlate some parameters and improve the overall accuracy. The in situ nature and the high flexibility of integrating different features of the calibration make the proposed method straightforward for most end-users.