September 26, 2020
Growth of commercial satellite imagery systems and data use are exploding in scope. At the annual JACIE conference over 200 dedicated remote sensing experts from industry, government and science came together to tackle the hardest problems in harmonizing use and standards for these incredible resources. If you missed this dynamic conference, here is a perspective on the key takeaways from this outstanding work.
1. There are major trends taking place in the technology and products for commercial imagery.
Data outputs are set to increase at a high rate, far beyond what human beings can process. Therefore, Machine Learning (ML) and Artificial Intelligence (AI) will be core components of future image delivery systems. Meanwhile, data consumers are demanding higher revisit frequency on target sites and low latency capability in returning useable data.
The general overriding trend in EO instrument development is to move towards higher resolution imagery.The industry is also working to develop multi-angle (3D) capability through agile satellite platforms.But it’s clear the next systems will be designed as instrument constellations, virtual constellations or “systems of systems” to meet these new products and market demands.
2. There is a growing need and importance of Analysis Ready Data (ARD)
Analysis Ready Data is the ability for data from any imager to be accurate and dependable for use in the end user’s application. “Analysis Ready Data (ARD – commercial)” or “Exploitation Ready Data (ERD – government)” and “Interoperable” satellite systems are a major objective for remote sensing. The lack of these standards is also is a limit to the current effectiveness of remote sensing data being generated today.
A key question in trying to create ARD standards is: “ARD for what?” Different applications and uses demand different levels of what “ARD” means and what needs to be done to make the data useful.
CEOS, NASA, ESA, CNES and other organizations have established good baselines for Analysis Ready Data standards but much more work is needed. There is a short timeline and high interest from the science, government, and industry stakeholders to develop ARD standards. But a unified approach to these three unique market segments is difficult to achieve.
3. Artificial Intelligence (AI) and Machine Learning (ML) will significantly augment if not replace the human analysis role in assessing commercial imagery.
There is more than enough data generated every day from commercial imagers to occupy all human analysists for hundreds of years. ML and AI must be developed to reduce this burden. The next generation of products need to deliver adaptive imaging systems that can anticipate where images will be needed without human intervention through Artificial Intelligence (AI).
A key question is, “What does a remote sensing satellite system look like when it was designed for AI exploitation rather than a human analyst?”
4. Calibration and verification techniques need to improve to support the market trends and development of useable ARD data sets.
Image quality today is hampered by limited calibration events, difficult techniques, and a lack of common methodology. Calibration and satellite performance evaluation techniques often differ significantly between high- and low-resolution images. Consequently, AI and ML outputs are only as good as the data that is fed into these programs. Therefore, without fundamentally better data, results may be easier to get, but still limited in scope of use.
5. Non-EO imagery such as SAR, thermal, microwave and GNSS are essential tools to compliment today’s EO imagery.
Standard imagery suffers from atmospheric effects. Alternate methods have proven useful to see through clouds to provide data other EO systems cannot. But interoperability and ARD is needed here as well. Data “fusion” products of EO and non-EO data is becoming an emerging need.
At Labsphere we have been actively engaged in understanding and creating innovative solutions to enable the industry to meet its’ future challenges.
SEE HOW LABSPHERE IS HELPING TO SOLVE THESE PROBLEMS WITH THE NEW FLARE CALIBRATION NETWORK: www.flare-network.com