Automated Detection of Spatial Distortions of Hyperspectral Push-broom Sensors - Application to the EnMAP Mission
Rogass, Christian; Brell, Maximilian; Segl, Karl; Kuester, Theres; Kaufmann, Hermann
Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences, GERMANY
The EnMAP imaging spectrometer utilises the pushbroom technology and senses the surface line-by-line using two separate scan lines for each detector. Attitude variations, micro vibrations and deviations from an ideal projection of incident light such as keystone cause variations of the line of sight of each detector element. These effects need to be reduced by the ground segment geometric pre-processing chain and to be validated over time. In this work an automated image based approach will be presented as potential tool for the geometric validation of the EnMAP products. It primarily focuses on the determination of the accuracy of band-to-band and detector-to-detector registration. For this, sub-pixel precise tie points are automatically selected based on an adaption of the Structure Invariant Feature Transform (SIFT) algorithm. These points are used to model the spatial distortions as high order polynomials for each band and detector. As a special feature of the proposed approach the results are additionally validated by an internal two-step technique of higher accuracy. First, the polynomial models are inverted and applied to reduce the spatial distortions. Second, the corrected data is then validated by inspecting detected spatial shifts using a phase correlation technique which is tie point independent. Conducted tests with simulated EnMAP data incorporating varying noise scenarios showed that spatial distortions can be detected of an accuracy of up to 1/100 pixel on average and internally validated of up to 1/1000 pixel on average. The proposed approach is generic, fully automated and has a high potential to support the geometric validation of other push-broom sensors such as SENTINEL-2.