In 1972 the crew of the Apollo 17 moon mission turned their attention back to Earth and took what would become the most reproduced photograph in human history.
The Blue Marble captures a picture of Earth from a distance of 29,000 kilometers, the hemisphere fully illuminated by the sun, stretching from the Mediterranean coast to Antarctica.
This was the first time humankind had seen our planet as a whole - within our lifetimes a picture of planet Earth was still a marvel. The Blue Marble helped spark the environmental movement, and came to symbolise the new era of globalisation.
It was also the last time a photo taken by human hands would show the full picture of our planet.
How does our planet look through the eyes of machines?
Today, some 1,100 satellites provide a picture of Earth the detail of which is without precedent in human history. Google Earth would have been considered a miracle only 20 years ago. Now anyone with a smartphone can see any location on Earth.
But the picture of our planet revealed by Deep learning and Geographic Information System (GIS) goes far further - and deeper - than the human eye can see.
Satellites provide images across the electromagnetic spectrum that reveal deep hidden secrets in the earth. Aerial monitoring, and now ultra low cost drones, can monitor huge areas to show detailed changes over time. Video camera networks collect human movement trends that show our cities as we have never seen them before.
The picture of Earth that machines reveal is already driving decision making in government and industry. Now that picture is set to become even more integral, as machine learning reveals new and deeper insights.
The fastest growth area in geospatial data analysis is machine learning, and the deep insights it yields.
Geospatial data is, to use a colloquial term, “heavy”. It is composed of high resolution images and video, and other data types that place huge demands on available computer processing power.
But the real challenge geospatial presents is in its analysis. Geoanalytics - the process of drawing meaningful information from such massive data sets - is a non-trivial challenge.
Human photo interpretation can be applied to the problem, with great success, via outsourcing solutions like Ingedata. But the fastest growth area in geospatial data analysis is utilising the fast growing discipline of machine learning.
Transportation & Logistics
Industries like petrochemicals and mining are faced with a core problem; how to implement change monitoring operations over thousands of square kilometres? Today, inexpensive drones and multiple satellites can provide near constant surveillance, but it’s machine learning algorithms that turn geospatial data into actionable intelligence.
Renaud Allioux, co-founder of Earthcube explains "We take dozens, potentially hundreds, of site and use AI and satellite imagery to track strong and weak signals for security, defense and financial markets."
A 7.2 percent reduction in fuel use is significant in many business contexts, but for Mexican trucking giant Coppel it represents an important step in the struggle to reduce CO2 emissions. The company is employing machine learning to discover which of their vehicle fleet are most efficient on different routes, revealing otherwise invisible energy saving patterns.
Where to position a new brand name storefront in a busy urban retail environment? The combination of video monitoring, customer loyalty systems and AI analysis can offer sometimes counterintuitive answers. The ability to track thousands or millions of consumers, to study how their brand affiliations interact, shows patterns of behaviour that a simple footfall study would never reveal.
In the face of tough competition from e-commerce, high street retailers are using geospatial machine learning not just to select where to open - or close - new stores, but to monetise the cross-channel value of their storefronts to specific brands. The rates brands pay to place products in prime locations is significantly higher when supported with compelling geospatial data.
No industry in the world is more reliant on data analysis than financial investors in the form of hedge funds and venture capital. The entire investment model is predicated on the certainty that data driven insights can reveal profitable investment opportunities before anybody else sees them.
AI analysis of geospatial data is at the forefront of investment decisions today. Not least, in cracking the secrets of other investors. It’s near impossible to make any major financial investment without building out infrastructure, moving equipment, people, or resources. Machine analysis of geospatial data can reveal the early stage of investment activity to competitors far in advance, letting investors get in on the latest opportunities.
Machine learning is already opening areas of tremendous potential in healthcare. In a spate of recent trials, AI diagnostic systems have exceeded the skills of human doctors in identifying eye diseases and many types of cancer. The same techniques of deep learning, applied to geospatial data mining, have exciting possibilities in the discipline of epidemiology: the study of behavioral and geographic influences on human health.
A recent study of air pollution in Los Angeles combined data from multiple specialised medical sources with big data from Open Street Map, fed into two tiers of the random forest learning algorithm, to predict concentrations of PM2.5 particles across the city. This is the kind of outcome that allows healthcare providers to accurately guide the behaviour of patients for long term preventative medical outcomes.
Geospatial artificial intelligence represents what must be considered a paradigm shift in how we understand both planet Earth and human life on the planet. Machines see our world with a depth and detail that humans may never truly understand - but that businesses must.
The businesses that capitalise on the potential of GeoAI, who understand the world as machines reveal it to us, will establish the new models of industry that will define the century ahead.