using cloud motion to calibrate webcams @ MVRL

Overview

We propose cloud motion as a natural scene cue that enables geometric calibration of static outdoor cameras. This work introduces several new methods that use observations of an outdoor scene over days and weeks to estimate radial distortion, focal length and geo-orientation. Cloud-based cues provide strong constraints and are an important alternative to methods that require specific forms of static scene geometry or clear sky conditions. Our method makes simple assumptions about cloud motion and builds upon previous work on motion-based and line-based calibration. We show results on real scenes that demonstrate the effectiveness of our proposed methods.

Citation

Nathan Jacobs, Mohammad T. Islam, and Scott Workman. Cloud Motion as a Calibration Cue. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013 [PDF] [Code]

@inproceedings{jacobs13cloudcalibration,
  author={Nathan Jacobs and Mohammad T. Islam and Scott Workman},
  title={Cloud Motion as a Calibration Cue},
  booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year=2013}

Highlights

Constrained Differential Flow

A key component of our approach is estimating cloud motion for every sky pixel in a video. We assume that the motion at each pixel is essentially constant over the video. We estimate two types of motion: one with and one without assuming a single translational cloud motion. For the case of a single translational motion, a red dot shows the location of the vanishing point of the motion. This point is the projection of the wind direction vector that is causing the clouds to move.

Correcting for Radial Distortion

Many outdoor webcams have significant radial distortion, therefore we apply a streamline estimation algorithm to our unconstrained flow estimates to detect lines. This figure shows examples of days with good, bad and really bad flow estimates. For the good day, if there was no radial distortion the lines should be straight in the image. We filter out the "bad" lines and use Prescott's method to estimate radial distortion.

Final Calibration Results

We combine estimates of cloud-motion vanishing points from multiple days to estimate the orientation and field of view of the camera. The figure above shows example final calibration results for the cameras we used in this work.

Code

Coming soon.

Acknowledgements

We thank Robert Pless, Austin Abrams, Jim Tucek and Joshua Little at Washington University in St. Louis for collecting most of the videos we used in this work. We also thank Jim Knochelmann for collecting and organizing the wind velocity metadata and for Weather Underground for making their weather archive available to us. This work was supported by DARPA CSSG (D11AP00255).

Copyright

The documents contained in these directories are included by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without explicit permission of the copyright holder.

Multimodal Vision Research Laboratory

MVRL