Refining the Interior Orientation of a Hyperspectral Frame Camera With Preliminary Bands Co-Registration

Abstract

Lightweight hyperspectral sensors carried by unmanned aerial vehicles (UAVs) are becoming powerful remote sensing tools for several applications, for example, forestry and agriculture. Sequential frame acquisition by scanning the spectral bands with tunable Fabry-Perot interferometer (FPI) is one of the technologies suitable for these applications. The accurate co-registration of the individual bands to produce a hypercube and the bundle adjustment of all bands are still challenging tasks. Because of the geometry and internal optical components of this kind of camera, modeling of the interior geometry of the image bands requires more than a single set of interior orientation parameters (IOP). This paper developed a new method that applies a preliminary two-dimensional (2-D) geometric transformation to co-register all bands, based on projective parameters estimated during the calibration process. This preprocessing avoids the use of several sets of IOPs, simplifying the computation of image orientation with bundle adjustment. Experiments using a close range calibration setup and a UAV-based aerial image block showed that the new method was effective and improved the accuracy of the three-dimensional (3-D) point determination. Accuracy of one times ground sample distance (GSD) in horizontal coordinates and 1.2 GSD in height coordinate was achieved in the bundle adjustment using a single set of IOPs.