Refining the Geometric Calibration of a Hiperspectral Frame Camera with Preliminary Bands Coregistration

Abstract

Lightweight hyperspectral sensors carried by unmanned aerial vehicles (UAV) are becoming powerful remote sensing tools for several applications, e.g., forestry and agriculture. Sequential frame acquisition by scanning the spectral bands with tunable Fabry-Perot interferometer is one of the technologies suitable for these applications. The accurate coregistration of the individual bands, to produce a hypercube, or the bundle adjustment of all bands are still challenging tasks. Because of the geometry and internal optical components of this kind of camera, the interior orientation of the image bands requires more than one single set of parameters. An alternative that is assessed in this paper is to apply a preliminary 2D geometric transformation to coregister all bands, based on parameters estimated during the calibration process. This pre-processing avoids the use of several sets of interior orientation parameters simplifying the steps of computation of image orientation with bundle adjustment. Experiments with calibration in a close range calibration set up and with UAV-based aerial images are presented showing the effectiveness of this approach.