| NSF Org: |
OCE Division Of Ocean Sciences |
| Recipient: |
|
| Initial Amendment Date: | October 31, 2016 |
| Latest Amendment Date: | October 26, 2017 |
| Award Number: | 1700923 |
| Award Instrument: | Standard Grant |
| Program Manager: |
Kandace Binkley
kbinkley@nsf.gov (703)292-7577 OCE Division Of Ocean Sciences GEO Directorate for Geosciences |
| Start Date: | November 1, 2016 |
| End Date: | October 31, 2018 (Estimated) |
| Total Intended Award Amount: | $97,822.00 |
| Total Awarded Amount to Date: | $117,336.00 |
| Funds Obligated to Date: |
FY 2018 = $19,514.00 |
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
615 W 131ST ST NEW YORK NY US 10027-7922 (212)854-6851 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
61 Route 9W Palisades NY US 10964-8000 |
| Primary Place of
Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | OCEAN TECH & INTERDISC COORDIN |
| Primary Program Source: |
01001819DB NSF RESEARCH & RELATED ACTIVIT |
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.050 |
ABSTRACT
Since November 2015 the Ocean Observatory Initiative (OOI) high-definition camera (CamHD) has been collection video clips of the deep water hydrothermal vent on the Juan de Fuca Ridge located off the coasts of the state of Washington. The CamHD is being used to study the change of the vent over time and flow of fluids. Scientists hope to use the time-series images to study animal behavior and changes in animal and microbial colonization associated with changes in fluid flow, temperature and chemistry in response to seismic and volcanic events. The data is all publicly available but the huge size of the archive and the lack of tools for easy retrieval make it difficult for investigators to take advantage of this unique data set. With this project the PIs hope to demonstrate the scientific value of the OOI CamHD system by providing tools to provide better access to the video data. This project has the potential to enable the scientific investigations in biological, geological, and oceanographic investigations that are yet to be imagined. This project will support two early-career scientists engaging in an interdisciplinary effort to add substantial value to this community resource.
At the time of the writing of this proposal some 33 TB data from the OOI seafloor high-definition camera (CamHD) has been collected and archived. The data is all publicly available but the huge size of the archive and the lack of tools for easy retrieval make it difficult for investigators to take advantage of this unique data set. The PIs have requested EAGER funding to demonstrate value of the proposed cloud-capable system for this data.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
he NSF-Funded Ocean Observatories Initiative Cabled Array (OOI CA) provides an unprecedented level of power and bandwidth for exploration of the deep ocean. The always-on connectivity provided by the cable enables new types of ocean instrumentation, such as the High Definition camera (CamHD) installed in the Ashes hydrothermal vent field, 1500m below the ocean's surface. Eight times a day, CamHD broadcasts a live stream of the life and geology of Mushroom, a ~2m tall hydrothermal chimney formed by hot, mineral-laden water emerging from the seafloor. Each time it broadcasts, CamHD looks over the entire western side of Mushroom, zooming in at a number of key locations to observe the vent in detail.
In addition to the live stream, each broadcast is archived in the OOI Cyber-Infrastructure (CI) as both a high-definition and a compressed (lossy) video. Due to the size and number of these files, it would be time consuming and expensive for an individual researcher to download and store the entire video time series.
This project supported the development of three software tools for accelerating the analysis of the archived CamHD video. The first is a tool for retrieving individual frames from the HD video files hosted at the OOI CI without having to download the full video files. This tool is designed as a web microservice, making it simple to retrieve individual frames from any web browser, or to build tools using any programming language with an HTTP client library.
The second tool runs motion and image analysis algorithms on every CamHD video, identifying time periods during which the camera is pointed at known sites on Mushroom vent. Since the camera runs a pre-programmed sequence, it views the same sections of the vent during every video, however, there's no reliable record of when the camera is looking at each site. Due to small variations in timing, particular camera positions don't necessarily occur at the same time within each movie.
This analysis occurs in two stages. For each video, an optical-flow algorithm is used to estimate the camera motion throughout the video and to identify time ranges where the camera is not moving (static regions). A second pass is used to label each static region with a scene identification tag by comparison with a set of manually-labelled ground truth files. The end result is a list of timeframes within each video where the camera is still, each labelled with a scene ID which is consistent across all videos.
The ID can be used to find and isolate video subsets or still images in all labelled CamHD videos which show the same location on Mushroom. For example, this allows creation of time-lapse videos showing the long-term evolution of hydrothermal structures on Mushroom, and it is also a precursor for further automated analyses of CamHD video.
This meta-record of labelled static scenes is published publicly (https://github.com/CamHD-Analysis/CamHD_motion_metadata) along with documentation on its calculation and a simple Python library for accessing video frames based on the meta-record.
The third component was the development of tools for accelerating the analysis of CamHD video files on both private and public cloud resources. Analyzing a single CamHD video requires about an hour on a high-end desktop computer, but the processing of each video is independent of the other videos. However, the project faced a multi-year / 50+TB backlog of CamHD video files and a continuous stream of new files arriving every day. To clear this backlog, large numbers of virtual servers were "rented" from the Google Cloud Platform over a period of approx. three months, allowing high-throughput processing of the entire CamHD archive. Once the backlog was cleared, ongoing processing of new CamHD files is handled by a smaller cluster of physical machines. The capacity to inexpensively leverage massive, temporary compute assets will be essential to performing any future cross-archive analysis.
Work derived from this project was presented in a number of venues, including the 2016 and 2017 AGU general meetings, the 2018 AGU Ocean Sciences Meeting, the 2017 NOAA Ocean Exploration Forum, and the 2016 and 2017 IEEE/MTS Oceans conference. It was also used for a tutorial/demonstration at the 2018 Cabled Array Hackweek held at the University of Washington.
Last Modified: 03/07/2019
Modified by: Timothy J Crone
Please report errors in award information by writing to: awardsearch@nsf.gov.
