Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

In Flight Boresight Rectification for Lightweight Airborne Pushbroom Imaging Spectrometry

Julien Yuuki Burkhard, Jesse Ray Murray Lahaye, Laurent Valentin Jospin, Jan Skaloud

TL;DR

This work tackles the challenge of geo-referencing lightweight push-broom hyperspectral imagery by developing an automatic pipeline that derives tie points and calibrates the boresight using only raw spectral data and potentially imperfect GPS/INS trajectories. It introduces a Bayesian model to estimate inter-line horizontal shifts with sub-pixel accuracy, coupled with a y-scale invariant feature matching strategy to handle pitch- and speed-induced distortions. Tie points are extracted using A-KAZE on rectified data and filtered via a homography-based RANSAC approach, enabling robust boresight estimation by minimizing epipolar constraints expressed on the lie algebra of rotations. Bootstrapped evaluations show that the proposed method improves horizontal rectification, increases inlier tie points, and reduces boresight uncertainty by substantial margins, approaching the accuracy of manual calibration and outperforming state-of-the-art automatic methods.

Abstract

Hyperspectral cameras have recently been miniaturized for operation on lightweight airborne platforms such as UAV or small aircraft. Unlike frame cameras (RGB or Multispectral), many hyperspectral sensors use a linear array or 'push-broom' scanning design. This design presents significant challenges for image rectification and the calibration of the intrinsic and extrinsic camera parameters. Typically, methods employed to address such tasks rely on a precise GPS/INS estimate of the airborne platform trajectory and a detailed terrain model. However, inaccuracies in the trajectory or surface model information can introduce systematic errors and complicate geometric modeling which ultimately degrade the quality of the rectification. To overcome these challenges, we propose a method for tie point extraction and camera calibration for 'push-broom' hyperspectral sensors using only the raw spectral imagery and raw, possibly low quality, GPS/INS trajectory. We demonstrate that our approach allows for the automatic calibration of airborne systems with hyperspectral cameras, outperforms other state-of-the-art automatic rectification methods and reaches an accuracy on par with manual calibration methods.

In Flight Boresight Rectification for Lightweight Airborne Pushbroom Imaging Spectrometry

TL;DR

This work tackles the challenge of geo-referencing lightweight push-broom hyperspectral imagery by developing an automatic pipeline that derives tie points and calibrates the boresight using only raw spectral data and potentially imperfect GPS/INS trajectories. It introduces a Bayesian model to estimate inter-line horizontal shifts with sub-pixel accuracy, coupled with a y-scale invariant feature matching strategy to handle pitch- and speed-induced distortions. Tie points are extracted using A-KAZE on rectified data and filtered via a homography-based RANSAC approach, enabling robust boresight estimation by minimizing epipolar constraints expressed on the lie algebra of rotations. Bootstrapped evaluations show that the proposed method improves horizontal rectification, increases inlier tie points, and reduces boresight uncertainty by substantial margins, approaching the accuracy of manual calibration and outperforming state-of-the-art automatic methods.

Abstract

Hyperspectral cameras have recently been miniaturized for operation on lightweight airborne platforms such as UAV or small aircraft. Unlike frame cameras (RGB or Multispectral), many hyperspectral sensors use a linear array or 'push-broom' scanning design. This design presents significant challenges for image rectification and the calibration of the intrinsic and extrinsic camera parameters. Typically, methods employed to address such tasks rely on a precise GPS/INS estimate of the airborne platform trajectory and a detailed terrain model. However, inaccuracies in the trajectory or surface model information can introduce systematic errors and complicate geometric modeling which ultimately degrade the quality of the rectification. To overcome these challenges, we propose a method for tie point extraction and camera calibration for 'push-broom' hyperspectral sensors using only the raw spectral imagery and raw, possibly low quality, GPS/INS trajectory. We demonstrate that our approach allows for the automatic calibration of airborne systems with hyperspectral cameras, outperforms other state-of-the-art automatic rectification methods and reaches an accuracy on par with manual calibration methods.
Paper Structure (14 sections, 11 equations, 6 figures, 2 tables)

This paper contains 14 sections, 11 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Prior works
  3. Methodologies
  4. Datasets
  5. Horizontal shifts and probabilistic image formation models
  6. Vertical scale invariance
  7. Tie point extraction and filtering
  8. Hyperspectral camera boresight calibration
  9. Experiments and evaluation
  10. Horizontal shift correction
  11. Tie points detection and filtering
  12. Camera boresight calibration
  13. Conclusion and future work
  14. Acknowledgements

Figures (6)

  • Figure 1: Overview of our proposed process for tie-point extraction in push-broom hyperspectral imagery
  • Figure 2: Dataset acquisition in Delemont (CH), map background © OpenStreetMap contributors
  • Figure 3: Probabilistic graphical model, as defined in Buntine_1994, used for the estimation of the horizontal shift. Blue circles are observed variables, white circles latent variables and dashed circles deterministic functions
  • Figure 4: The effect of y-scaling, where the aspect ratio of the image changes when the airborne platform flies parallel (a) or perpendicular (b) to the road.
  • Figure 5: Horizontal shift error distribution (according to to $\Delta x$ predicted by GPS/INS)
  • ...and 1 more figures