Google’s “Ground Truth” project is responsible for creating all the mapping data used by the company’s products. The Atlantic gained exclusive access to the company to see the process in action, and the resulting news article explains in detail how data is collected, verified, and published.
Google uses a variety of sources to collect its mapping data. Partners supply satellite imagery but a computer, of course, cannot understand the actual content – the roads, the junctions, or the buildings. To start turning satellite imagery into navigable maps, more data is collected from other sources, including national censuses and the USGS. These two data sets are “layered” together and start to build up a useful map. Of course, for various reasons the satellite imagery and road data do not always match.
What surprised me was how many of the minor tweaks to solve these problems are done manually – and how quickly changes can be made. Interestingly, one way the team verify the data is by cross-checking with the Google Street View team to see if their car has ever driven down that road – thus proving the road exists and is correctly placed.
One recent development is that as computer vision technology has improved, Google have been able to match famous business logos – thus being able to automatically recognise and map all Burger Kings, for example. And as OCR (OPtical Character Recognition) continues to improve, the Google Street View car will be able to “read” words all around it – potentially making them part of Google’s search index – for better and worse.
The Atlantic article is long but well worth a read, and contains a lot of information that should help ITGS students answer “step by step process” questions on the Paper 2 exams. Read full story on The Atlantic.