In recent years, a significant increase in private investments and robust government support in space are fueling rapid innovations and expansion of the ecosystem. Looking ahead, we believe the convergence of artificial intelligence (AI) and machine learning (ML) with other innovations is one of the most disruptive technology trends. AI is transforming the entire space ecosystem, from manufacturing to in-orbit operations tackling collision avoidance and optimizing spectral efficiency. AI on-board processing would enable more timely delivery of insights from space. Moreover, AI on the ground could optimize satellite mission control and management of mega-constellations. Most importantly, AI is enabling extraction of valuable insights from all the new geospatial data. This is transforming not just the space industry but is also revolutionizing global enterprises and government operations. Even the Pentagon has declared AI a national priority and is seeking partnerships with tech giants and startups to accelerate AI adoption across all operations.
Drowning in Data, But This is Just the Tip of the Iceberg
Until recently, space was dominated by governments using large and expensive exquisite satellites. But now the satellite industry has significantly lower barriers to entry with the onset of the small sat revolution which is driving an explosion of new data with higher resolution and revisit. In addition, growing drone fleets, Internet of Things (IoTs) and new sensors are collecting ever more data of the earth that it has become humanly impossible to manage, let alone analyze. With close to 1,500 new earth observation (EO) small satellites expected to be launched in the coming years, the explosion of data we are currently experiencing is just the beginning. This is in addition to the 400 plus small sats already launched by startups like Planet and Spire over the last five years.
New Sensors are Driving Even More Data and New Applications
New sensors such as Synthetic Aperture Radar (SAR), Hyperspectral, Infrared, Automated Identification System (AIS), and Radio Frequency (RF) Sensing are delivering new data sets to complement traditional optical imagery, enabling new commercial applications through advanced analytics. SAR, unlike optical, can capture imagery in all weather conditions, at night and through clouds. SAR startups include ICEYE, Capella, and Umbra Lab etc. Hyperspectral imaging can be used to find objects and identify materials with applications in oil and gas, mining, and agriculture. Startups building hyperspectral constellations are Satellogic, NorthStar Earth & Space and HyperSat. Infrared (IR) or thermal imaging is complementary to other sensors; SatelliteVU, ConstelIR and Koolock are building IR constellations for weather forecasting, environmental monitoring, and defense applications. Next, automated identification system (AIS), one of the payloads on Spire, which also provides weather and aircraft tracking data, allows maritime tracking and monitoring of vessels. RF Sensing startups HawkEye 360 and Kleos identify and geo-locate specific radio signals, which provide valuable insights on maritime awareness when ships have turned off their AIS tracking device.
Technology advances in computing power, cloud and AI/ML have enabled automated extraction of insights from massive data at scale. This is done by training the machines to mimic humans to perform repetitive tasks such as data collection, processing, object and change detection from satellite imagery. Geospatial AI analytics are also used to make predictions such as crop yields and weather forecast. More importantly, AI enables fusion of remote sensing data with other ground data sets (news and social media feed, terrain data etc.), creating big data analytics to solve large and complex problems from environmental monitoring, disaster response to enabling self-driving cars.
Customer Want More Insights and Not More Pixels
Traditional Earth Observation players like Maxar Technologies and Airbus, joined by EO startups such as Planet and Spire have all expanded beyond selling mere imagery to valued-added services based on AI. At the same time, a wide range of VC-backed geospatial analytics startups are leveraging lower cost and growing data sources to deliver geospatial intelligence or insights to enterprise and government customers. Customers are no longer interested in pixels only and look for providers to help them extract insights to support better decisions.
Leading geospatial analytics startups pivot to focus on Defense/Intel. Orbital Insight, Descartes Labs and Ursa Space Systems have each pursued a different strategy, but all have pivoted to focus on the Geospatial Intelligence (GEOINT) market for defense. These companies offer data agnostic platforms and deliver geospatial analytics as a subscription service. Orbital Insights uses Machine Learning and Computer Vision to identify economic trends at scale. For example, it provides automated trends analysis of cars in retailer parking lots, track global energy supply and predict crop yields for financial traders. Descartes and Ursa offer horizontal data platforms for customers to access and run analytics on pre-processed multi-sourced imagery. Ursa is focused on providing access to analytics ready SAR data.
Highlighting Three Startups Gaining Traction with Vertical Market Focus
First, Kayrros is a data analytics startup with deep domain expertise in the Energy market. Like Orbital Insights and Ursa, they sell energy indicators to traders, but they also deliver insights to the Oil and Gas industry by tracking global storage, transportation and production. Second, PlanetWatchers delivers all-weather data analytics for large scale natural resources and infrastructure monitoring by fusing high-resolution SAR with optical imagery. User cases include oil and gas pipeline monitoring, forestry drought and early disease detection. Lastly, Delair started out in drone inspection, but has expanded into Big Data Analytics targeting mining and construction. It is working with large global companies to accelerate their digital transformation by developing digital twins of their assets and activities using drone collected aerial imagery combined with satellite, sensor, and other data sources
In summary, AI/ML is critical in unlocking the value of remote sensing data. Geospatial AI analytics remains a nascent market and continues to face the challenge of inadequate training data and a shortage of talent. While there is an acceleration of startups, only a few new constellations are currently operational (Planet, Spire). In reality, there is often not enough high resolution, high frequency satellite data available to meet customer expectations, as well as for AI training to deliver highly reliable results. Lastly, AI will always require a human in the loop as it is as much of an art as science. Domain expert/user is key in working with data scientist to build the right models with the right data input, as well as to test and evaluate the algorithms and models to improve accuracy. Digital transformation occurs when enterprise can incorporate geospatial data/location intelligence in their workflow to drive significantly higher productivity.