This year’s breakthroughs in artificial intelligence (AI) capabilities have captured the imaginations of businesses, governments and the public alike.
Part of the reason AI has generated more widespread excitement than other recent tech advances, such as blockchain technology or the metaverse, is its sheer game-changing usefulness. AI has the potential to revolutionise a range of tasks, from planning and research to idea generation and content creation.
One real-world use that has already proven invaluable is the deployment of AI and machine learning to generate mapping – a sweet spot that combines the efficiency and utility of AI with the kind of geolocation technology that has already improved lives in countless ways, from satnavs and delivery apps to real-time transport updates.
Maps are essential for planning and managing infrastructure, transportation systems, land use, and efforts to mitigate environmental damage. But keeping a master map of a country up to date can be an expensive and laborious process. Even with modern tools, such as satellite photos or images taken by drones, the work to digitally trace vast urban and rural areas manually can be hugely time-consuming and expensive.
“Imagine the amount of buildings there are in one square kilometre in a dense urban environment – it can take even an experienced operator up to a week to manually capture all of those buildings,” explains Mark Tabor, a principal consultant at Ordnance Survey (OS), Britain’s national mapping service. Indeed, in Britain, OS has a team of 250 surveyors tasked with this job and with keeping the national mapping database up to date. But this is a luxury that not every country can afford.
Enter AI, which OS is employing in pioneering techniques for building detailed digital maps quickly and accurately, and at low cost. The technique, which OS is using in Britain to improve efficiency, essentially involves teaching computers what to look for in images, using training data so that maps can automatically be created.
One striking example of how its cutting-edge techniques have been put to use is in Lusaka. The capital of Zambia has undergone rapid urbanisation in recent years – its population has more than tripled from a million at the turn of the millennium to an estimated 3.2 million people today. That transformation has been a huge challenge for the Zambian officials tasked with running public services and managing the city – a challenge compounded by the lack of an accurate, up-to-date map, which is one of the crucial building blocks of any modern city. Without a map, officials couldn’t be certain where and how people were living, or the services they required. And this was particularly important in a city where many citizens live in informal settlements that haven’t gone through any kind of planning processes.
“When the local government thinks about how it’s going to organise public transport, it needs to know not only where the informal settlements are, but how [residents] actually get to their place of work,” says Tabor, who has been working with the Zambian government. “The mapping the government had was out of date.” This meant that before its new map was created, it was having to rely on other sources, such as Google Maps and OpenStreetMap – a geographic database maintained by volunteers. But these are not accurate enough for mission-critical government work. “They’re not always up to date, they don’t have a consistent specification, and it means that when there are some really crucial pieces of information that the government needs to derive from mapping, it doesn’t have the latest data,” says Tabor. “These other sources might be great for navigation and understanding where the shops are, but, if you want every single property to not only be on your map, and also to have additional attribution, not so much.”
He gives the example of how a detailed map of every dwelling can be essential in disaster situations such as flooding. If the government has reliable location data, it can match homes to residents, so that emergency services can figure out how many people may require help, and where they are likely to be located.
To ensure the accuracy of OS’s AI-powered base map, the starting point was up-to-date aerial photographs of Lusaka. “We began classifying all of the pixels into buildings and roads, and other different types of features that were asked for by the end user.” says Tabor. What the surveyors were doing was training the system to recognise things. Each time a box was drawn around a home, or a line was traced over a street, it taught the computer how to spot these sorts of features on its own, without human help. And the more training data that was fed in, the better the automated end results. “It’s a balance between having a fully automatic process where you don’t need any training, which might get to, say, 85% completion, or by investing in more training data for even greater accuracy,” says Tabor.
Using this technique, Tabor’s team was able to automatically map the entire 420 sq km of Lusaka far more rapidly than if it had been done by humans. They also carried out checks on the work of the system, as there will always be some unusual buildings that the AI might classify incorrectly.
The AI-generated maps are also highly versatile. They’re able to keep up with rapidly growing urban areas or changing landscapes. They also allow for the integration of other sources of data, such as census data. Indeed, last year, Zambia completed its most recent census using the new maps.
Early in his research into this technique, Tabor discovered that the AI model was also more than capable of working across borders. It’s now a critical tool for OS as it helps other nations plan and manage their growth and development.
The organisation is already applying its knowhow to projects that range from identifying parcels of land for tax authorities, to using AI to classify different wildlife habitats. Tabor points to how accurate mapping data can also underpin public health policy: “If you’ve got to think about where you are going to have an inoculation centre, do you actually understand where the majority of the population is? How many inoculations are you likely to need?”
AI-powered mapping has also helped OS to automate the production of climate-related location data. OS has worked with Deimos Space UK and the Mohammed Bin Rashid Space Centre in Dubai to use AI and satellite data to identify and monitor palm trees and mangroves – types of vegetation that can help mitigate the impact of climate change by storing a large amount of carbon, and which are being planted in swathes.
Robust and reliable location data is also important when trying out new and innovative approaches to urban planning – for instance, the development of smart-city initiatives that seek to use data and digital technology to manage cities better. OS is working with Dubai on its smart city strategy, having previously worked with Singapore in a similar capacity.
For instance, Dubai required an up-to-date approach to capturing location data in order to create a so-called happiness meter, a system for gathering feedback from people across the city on everything from geographic areas to their experience as a customer. The happiness meter forms part of Dubai’s broader smart-city initiatives, and it turned to OS to undertake a geospatial readiness assessment to help understand the city’s needs and capabilities, and to identify the next steps that Dubai municipality needed to take.
Smart-city initiatives can allow authorities to go beyond service and infrastructure provision in their efforts to improve the lives of residents. Another example is Singapore, where OS helped smart-city initiatives by working with the National University of Singapore and the government by improving data interoperability and 3D data modelling to help the country plan its future more effectively. Digital models, sometimes referred to as digital twins, can help cities plan and evaluate interventions more effectively.
“There are also a lot of countries that are highly developed and have their own version of a digital twin, if you like,” says Tabor. “Singapore is a very good example.”
It’s just another way that innovative mapping techniques have become vital for innovative approaches to development, urban planning, and countless other challenges faced by public authorities.
See a sustainable place | Let OS shine a light on your world | Ordnance Survey
