Commercial remote sensing experts agree: the community is in the midst of a revolution
Commercial remote sensing experts agree: the community is in the midst of a revolution.
According to Robbie Schingler, co-founder and chief strategy officer of Planet, launch is the single greatest barrier to innovation in the remote sensing industry. Schingler voiced his support for secondary payload offerings such as cubesat deployer ISIPOD and NASA’s NanoLaunch, a program for small, low-cost vehicles to deploy cubesats in space.
Just three months ago in India, 104 satellites were launched on a single rocket, shattering the world record three times over. Planet owned 88 of those satellites. When fully deployed at the end of this month, Planet’s 150-satellite constellation will be able to function as a line scanner, imaging the entire surface of the earth every day.
Schingler also highlighted the sensor revolution on the ground, the need for open space situational awareness and traffic management protocol (similar to traditional air traffic management), and Earth observation trends such as pixel pattern analysis and machine learning.
“GED Talk” Perspectives from Space: Big-Sat, Small-Sat, and Clustered from Trajectory On Location on Vimeo.
Abe Usher, CTO of DigitalGlobe Radiant, said as remote sensing data increases in volume, machine learning and automation will become necessary to consume such large swathes of imagery. Usher discussed how proxy variables are being used to improve computer vision and automated perception of digital images. Pizza deliveries to the White House and Pentagon, for example, can indicate high-intensity national security planning and possible subsequent military action, as was the case in the nights leading up to the launch of Operation Desert Storm.
Modern machine learning algorithms can train computers to learn a level of human-like subject matter expertise and detect similar contextual clues.
“The technology is not perfect, but it’s maturing. If we can do this for on-the-ground photos, why not do this with video?” Usher said. Usher isn’t the first to pose such a question; off-the-shelf algorithms such as YOLO are mature enough to detect objects within video content faster than real-time.
DigitalGlobe is working alongside CosmiQ Works and NVIDIA to advance this sector of the tradecraft with an open data initiative called SpaceNet, which releases pre-labeled overhead imagery models into the public domain so others can create their own machine learning models for object identification.
That collaborative framework isn’t exclusive to DigitalGlobe. Other remote sensing players like Spaceflight Industries have also recognized the necessity of open-source sharing if the industry wants to progress as fast as the tech it relies upon. Spaceflight’s BlackSky imaging platform provides users access to internal satellite media as well as social media, traditional news media, and the user’s proprietary material. For example, BlackSky Events—one of the platform’s analysis offerings—used machine learning to gather relevant, publicly available text data and visuals about the June 3 London attack within a matter of minutes.
According to BlackSky Constellation Operations Engineer Chantz Thomas, the company’s other key offering is BlackSky Spectra, which provides automated feature extraction in addition to a web-based database of imagery from Spaceflight and other partners.
Posted in: GEOINT Foreword, GEOINT Symposium Tagged in: 2017 Show Daily Day 3, Machine Learning & AI, Remote Sensing, Small Sats
Multipurpose Science for a Changing World
David Applegate, Ph.D., Director of the United States Geological Survey (USGS), gave closing keynote remarks on the final day of USGIF’s GEOINT 2023 Symposium in St. Louis, Missouri.
GEOINT Symposium, Government Hub
NGA’s Mark Munsell Envisions 100% Accuracy with AI
The NGA Director of Data and Digital Innovation predicts success with computer vision
GEOINT Symposium, Keynotes
DNI Haines: GEOINT ‘Fundamental’ to U.S. National Security
Director of National Intelligence Avril Haines discusses the improvement and expansion of GEOINT—and where we go from here