A big data issue for DOTs is, “how do we extract data from design when it is minimally attributed?”.

GDOT’s Office of Traffic Data approached us to help them solve this problem. Last year we tackled took on the challenge using CAD APIs. With CAD, we were able to extract data with some success. The CAD programs however, had some problems. First, CAD applications are complex programs requiring a level of expertise most users, outside of design and engineering, do not have. Second, the properties associated with legacy drawings is minimally attributed and these properties are not always reliable. Finally, the transfer of data discovered from design to other disciplines, like GIS is cumbersome at best.

This year, we decided to approach the problem with digital twin technology. ​This decision facilitated the use of technologies like machine learning and AI. We have enabled our AI to detect road characteristics from design drawings and are using that data to generate attributed data that can then be consumed by GIS. Further, the use of a browser based digital twin means we are able to introduce an easy to use UI accessible by more users.

With our iTwin AI, we are able to detect turn lanes, count the number of lanes, extract merge/diverge features and more. The tools to accomplish this are more readily available using browser based technologies. And, we have the ability to add controls and features that make sense for a larger set of users. In the image to the right we show how the AI is able to identify turn symbols, group them into logical sets, and determine direction of travel.