Abstract
Since 2006, the Image Analysis and Data Fusion Technical Committee (IADF TC) of the IEEE Geoscience and Remote Sensing Society (GRSS) organizes a yearly Data Fusion Contest that aims to promote the use of new remote sensing data sources and stimulating new methodological developments [1]-[10].
Introduction
The 2016 IEEE GRSS Data Fusion Contest challenges the remote sensing community to bridge the gap between traditional satellite imagery and the emerging frontier of "video-from-space." While satellite imaging has traditionally focused on static "snapshots," the rise of space-based video platforms introduces a massive shift in temporal resolution. However, integrating these high-frequency video streams with traditional, calibrated multispectral satellite data remains a significant technical hurdle. The contest aims to spark novel methods for tracking, change detection, and multi-sensor integration by providing high-resolution multitemporal data alongside Ultra-High Definition (UHD) video captured from the International Space Station (ISS).
Contest Framework and Research Directions
As an announcement for a contest, this framework defines a "paper competition" where participants develop novel approaches rather than following a single prescribed algorithm. Key research directions encouraged by the contest include:
- Registration: Aligning video frames to static satellite images or vice versa.
- Object Detection and Tracking: Identifying and following moving targets like cars, vessels, or aircraft using the 3 frames-per-second (fps) UHD video.
- Change Detection: Using the multitemporal sequence to identify urban evolution.
- Multisensor Fusion: Combining 4-meter multispectral bands with 1-meter panchromatic and video data to create high-resolution, color-accurate products.
Multitemporal and Multisensor Datasets
The contest provides a unique dataset centered on Vancouver, Canada, featuring two distinct modalities:
- DEIMOS-2 Multispectral Sequence: Two sets of images (March and May 2015) providing 1-meter panchromatic data and 4-meter multispectral data (Red, Green, Blue, and Near-Infrared).
- UrtheCast Iris UHD Video: A full-color video captured from the Zvezda module of the ISS on July 1, 2015. It features a 1-meter ground sample distance at 3 fps, covering an area of roughly 3.8 × 2.1 km.
By providing Level 1C orthorectified data, the organizers have lowered the barrier to entry for fusion, allowing researchers to focus on high-level analysis rather than manual image alignment.
Innovation and Community Impact
The inclusion of ISS-based video marks a transition from monitoring static states to monitoring dynamics. Unlike many machine learning competitions that rank participants on a single leaderboard metric, this contest encourages researchers to define their own problems and find creative ways to exploit the complementary nature of the data.
Successful innovation is rewarded through publication opportunities at the IGARSS 2016 conference and in the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (JSTARS). The ultimate goal is fostering new methodologies for handling multitemporal and multisource data in the remote sensing community.
Limitations and Future Outlook
The potential of video-from-space is still being explored. Current limitations include the specific orbital constraints of the ISS and the high volume of data generated by UHD video. Future work involves moving toward fully automated pipelines that can ingest these disparate data streams to provide real-time insights into urban dynamics, environmental changes, and global logistics.
Summary updated