New Computer Vision Routine Developed for CAMHD Time-Lapse Video
A new computer vision routine, developed by Aaron Marburg at UW-APL and aided by Tim Crone at LDEO and Friedrich Knuth at Rutgers, is now able to correctly identify and tag scenes of scientific interest in the CAMHD video stream. These scenes were previously being manually identified by students at Rutgers University, a process which has been greatly accelerated by the team’s work. With this enhanced metadata record, a brand new set of time lapse videos has been created, displaying a frame captured every three hours from November, 2015 to July, 2016. There are 9 scenes of scientific interest, which are recorded at two or three zoom levels, depending on the camera routine. The naming convention for the videos is deployment_(dx) position_(px) zoom-level_(zx). For more information on the different scene tags, see the regions description on GitHub.
Position: 0
Zoom Level: 0
Zoom Level: 1
Zoom Level: 2
Position: 1
Zoom Level: 0
Zoom Level: 1
Position: 2
Zoom Level: 0
Zoom Level: 1
Position: 3
Zoom Level: 0
Zoom Level: 1
Zoom Level: 2
Position: 4
Zoom Level: 0
Zoom Level: 1
Zoom Level: 2
Position: 5
Zoom Level: 0
Zoom Level: 1
Zoom Level: 2
Position: 6
Zoom Level: 0
Zoom Level: 1
Zoom Level: 2
Position: 7
Zoom Level: 0
Zoom Level: 1
Position: 8
Zoom Level: 0
Zoom Level: 1
Resources
Raw Data Archive – Access to raw video data files in .mp4 and .mov format
Live Video Feed – Live video from Axial Seamount, every 3 hours for 15 minutes
CamHDAnalysis – Computer vision routine used to create metadata and time lapse videos
CamHD Compute Engine – Open access processing platform to use and develop CAMHD processing code
Download Time-Lapse Videos – Shared drive that permits download of the time-lapse videos displayed above
References
Crone, T.J., Marburg A., Knuth F. A, Using the OOI Cabled Array HD Camera to Explore Geophysical and Oceanographic Problems at Axial Seamount, presented at the 2016 Fall Meeting, AGU, San Francisco, CA
Knuth, F. A., Marburg, A., Crone, T. J. (2017) Deriving Quantitative Metrics from OOI High-Definition Video Data for the Purpose of Automated QA/QC, Oceans 2017 MTS/IEEE, Anchorage, AK, September 18-21
Knuth, F.A., Marburg, A., Belabassi L., Garzio, L., Smith, M., Vardaro, M. (2016) Automated QA/QC and Time Series Analysis on OOI High-Definition Video Data. Oceans 2016 MTS/IEEE Proceedings.
Marburg, A., Knuth, F. A., Crone, T. J. (2017) Cloud-Accelerated Analysis of Subsea High-Definition Camera Data, Oceans 2017 MTS/IEEE, Anchorage, AK, September 18-21
Funding
National Science Foundation, Cloud-Capable Tools for MG&G- Related Image Analysis of OOI HD Camera Video, Award #1700923
For additional information please contact: help@oceanobservatories.org