Geospatial scientists have tapped into Google Street View images and developed a new program to manage street signs that need replacement or repair. The scientists trained the fully-automated system with the use of AI-powered object detection to identify street signs in the freely available images. They published the result of their study in the journal of Computers, Environment, and Urban Systems.

At present, municipal authorities spend large amounts of time and money to monitor and record the geolocation of traffic infrastructure manually, a task which also exposes workers to unnecessary traffic risks. The results of the research revealed the system detects signs with near 96 percent accuracy, identifies their type with near 98 percent accuracy, and can record their precise geolocation from the 2D images.

According to Andrew Campbell, the leader of the study and RMIT University Geospatial Science Honors student, said that the proof-of-concept model was trained to see 'stop' and 'give way' (yield) signs, but could be trained to identify many other inputs and was easily scalable for use by local governments and traffic authorities. Explaining further, Campbell said municipal authorities have requirements to monitor this infrastructure but at present, no cheap or efficient way to do so.

Campbell noted that with the use of free and open source tools, they have now developed an automated system for doing that job, and doing it more accurately. During the reviews, the researchers discovered that mandatory GPS location data in existing street sign databases were often inaccurate, something up to 10m off.

Campbell said further that tracking these signs manually by people who may not be trained geoscientists introduces human error into the database. Once set up, their system can be used by any spatial analyst, tell the system which area to monitor, and it looks after it.

Dr. Chayn Sun, the RMIT geospatial scientist, noted that the fact that some councils were already attaching camera onto rubbish trucks to gather street footage showed how valuable visual data were becoming, given what technology could now do with it. Sun explained that this imagery is critical for local governments to monitor and manage assets, and with the vast amount of geospatial applications flourishing, this information will only become more valuable.

More like the footage from rubbish truck cameras or any other geo-referenced imagery of the road network collected by municipal authorities, Sun said it could also be fed into the system. She concluded that where footage is already being gathered, their research can provide councils with an economic tool to drive insights and data from this existing resource.