Navigating the Data Deluge in Geospatial Imaging: Challenges and Solutions
Despite our inherent vanity, humans have never scrutinized themselves as meticulously as we do now. In recent years, there has been an exponential surge in the number of both public and private satellites traversing various orbital routes. Currently, over 8,000 satellites orbit our planet, and we expect to have 58,000 by 2030. We keep tabs on everything – from weather patterns and ocean currents to farmlands, highways, ports, cities, and forests – there is always a vigilant eye in the sky.
Geospatial imaging is the monitoring and imaging technology that powers this new industry. Light sensors capture images similar to what we see with our eyes. These sensors observe the reflection of sunlight on objects, creating pictures that are both familiar and intuitive to understand. Yet, this is merely the tip of the iceberg when it comes to geospatial imaging technologies.
For instance, infrared (IR) imaging can offer a different perspective by capturing heat radiated from objects. IR proves incredibly useful in various fields, such as forestry, where infrared can identify hot spots in a forest that could signal an early-stage fire, or in agriculture, where it can reveal stressed crops before they are visibly impacted.
Other types of imaging include Synthetic Aperture Radar (SAR), which operates in the microwave spectrum and can see through clouds and even darkness, providing all-weather and all-day monitoring capabilities. SAR is particularly valuable for observing maritime activity or tracking changes in polar ice, where cloud cover and long nights can hinder visible light sensors.
Each imaging technique brings unique advantages, and the choice of technology depends on the specific requirements of the task. Some technologies provide broad overviews of large areas, and others can zoom in on minute details. Some are excellent at detecting certain types of materials or activities, while others offer a more general view of the world as we know it. The growing diversity in imaging techniques leads to an unprecedented depth and breadth of information about our world.
However, this proliferation of geospatial satellites and monitoring brings its challenges, namely just how much data is being sent back to earth each day. It's like having too many eyes and not enough brains to process what they're seeing.
Addressing this data deluge is driving the next wave of innovation in the geospatial industry. Tools like AI and ML can sift through vast amounts of data, identifying patterns and anomalies that human analysts could easily miss. As we move forward, the focus is shifting from simply collecting data to making it actionable. While there may be too much geospatial data to handle today, the solutions on the horizon promise a future where we can not only watch but truly understand our world in real-time.
Data Explosion
The overwhelming success of this industry is punctuated by the sheer volume of the data produced. Your phone’s photo quality has improved alongside camera technology, and the same goes for satellite images. To provide a sense of scale, consider imaging industry leader, Planet Labs. They generate 30 terabytes of data daily, and their archive stretches back six years. Although impressive, this vast quantity of data is just a collection of overpriced selfies if we can't glean insights from it. Generating this level of data is already remarkable enough, but getting it back down to our computers on the planet's surface is another challenge entirely.
Transmission is the literal bottleneck in geospatial imaging. Since hardwiring a cable to a satellite is impossible, contact with Earth is done through radio frequencies (RF), and RFs are of course constrained to the laws of physics. They can't travel faster than the speed of light, meaning there's a palpable delay in communication satellites that are thousands of miles above us, especially when transmitting data to and from distant geosynchronous satellites. The delay increases when data is relayed between multiple satellites. This issue is further compounded by the complex measures for security and privacy (encoding, transmitting, and decoding) involved in satellite communication.
The bottleneck gets worse when you look at the next generation of imaging satellites. Hyperspectral imaging is currently in the early stages of deployment by companies like Wyvern and Pixxel. This process collects data over a broad spectrum of light, resulting in a comprehensive, highly detailed picture of the Earth's surface – rendered in 3D. This type of imaging far exceeds conventional standards, creating an even greater volume of data than traditional imagery technology, and presenting a new challenge in data management and analysis.
One possible solution is to boost communication efficiency. The Optica Foundation, an entity advancing optics and photonics worldwide since 1916, is financing research in integrated photonics. With a profound commitment to accelerating scientific, technical, and educational achievements, Optica is blazing a trail beyond the boundaries of silicon by using light waves for data transmission and processing instead of conventional electrical currents.
There are companies working on this solution, as well. Skyloom offers high-capacity optical data transport services that improve speed and availability in low Earth orbit. This isn't just about data collection; it's about ensuring data can be received and acted upon promptly. Skyloom is addressing what they call the "dial-up era" problem of spaceborne communication. They have established secure, high-capacity bidirectional connections for satellite-to-satellite and satellite-to-earth transmissions. This means that from the point of data collection to the cloud, every link in the transmission chain is done at the speed of light, and with lasers.
However, the magnitude of the data produced poses another challenge: analysis. Sifting through each square of geospatial imagery would require an immense human effort, something that is not feasible on a large scale. Though traditional AI and machine learning tools can provide faster insights, we are limited by their current scope of ability and high energy price of operation and training.
One of the main hurdles lies in data labeling, a critical step in training traditional AI models for machine learning applications. This resource-intensive process demands substantial time, data, and human resources. The scale of the challenge is such that even the Pentagon's AI chief has said, "If we don't label at scale, we're not going to win."
AI to the Rescue
Satellites are getting smarter, built with more processing power and storage than ever before. Onboard AI, or AI on the edge, can process data on the satellite itself, enabling quicker decision-making and reducing the amount of data that needs to be transmitted back to Earth. AI on the edge departs from the traditional method of transmitting all of the raw data to the ground for processing and analysis - after all, a lot of the imaging out there ends up on the cutting room floor. Instead of worrying about footing the bill to get it back to Earth, then shelving it, why not only transmit what is most interesting? When something of import does appear, the full record can be pulled as needed.
An excellent example of this solution is the partnership between Satellogic and Palantir. In April 2022, they launched their first Edge AI-enabled satellite into space. This joint project involved Satellogic adapting its hosted payload program and edge computing hardware to run Palantir’s Edge AI platform. This collaboration results in reduced latency and allows for faster insight generation by starting data processing upon capture. Their onboard AI also optimizes bandwidth by making AI-driven decisions about which data will be the most valuable to customers.
Even if onboard AI becomes standard, AI/ML technologies will still need to play a key role on the ground in processing and analyzing this data. One of the companies leveraging AI to provide solutions in geospatial data is Synthetaic. Synthetaic excels at unstructured and unlabeled data. By generating detection or classification models for real-time insights, Synthetaic can get answers from satellite imagery that previously took days-to-months to analyze. The recent Chinese balloon incident was a great example of its application. Synthetaic used their proprietary AI to match objects resembling his balloon sketch in the provided satellite imagery, with the first satellite image of the balloon found within just two minutes.
The Future
As we navigate this new era of geospatial imaging, it's clear that there are substantial issues that make the most of emerging technologies. The volume of data generated is overwhelming and will only grow in size. However, there are examples of innovative solutions that show a path forward. From AI on the edge to integrated photonics and free-space optical communications, the advancements in data management, processing, and transmission, the rate of technological progress is promising. Companies like Synthetaic and Skyloom are revolutionizing the market and demonstrating the potential of these innovative solutions in practical scenarios.
This vast and rapidly increasing amount of geospatial data isn't just an overwhelming treasure trove of information; it's a potent tool for addressing the planet's most pressing challenges. The invaluable data generated from geospatial imaging is crucial in combating global issues like climate change. By accurately monitoring weather patterns, tracking the melting of polar ice caps, observing shifts in agricultural yields, and keeping tabs on urban expansion, we can develop more effective mitigation strategies. Furthermore, geospatial information can help optimize resource management, planning, and disaster response.
The importance of real-time, precise geospatial information cannot be overstated in an increasingly interconnected and dynamic world. As the use cases for this data broaden and deepen, our ability to effectively interpret and act upon this information will aid our understanding of the Earth and provide critical insights for shaping a sustainable future. The challenges we face in data management, processing, and transmission are significant. Yet, the potential rewards are vast - enabling us to transform how we respond, interact, and care for our planet.
The road ahead may be complex, but the potential benefits of these technologies make the journey well worth it.
