Researchers at Dartmouth College and the University of Bologna in Italy have developed an Android app that uses the camera on a smart phone to detect oncoming traffic.

The app relies on machine-learning and image-recognition algorithms to identify the fronts and backs of vehicles, and takes into account varying light conditions, phone tilt, and blur. When WalkSafe detects a car approaching at 30 miles per hour or faster, it vibrates the phone and makes a sound to alert the distracted user.

Andrew Campbell, professor of computer science at Dartmouth and head of the Smartphone Sensing Group, says the app also exploits phone APIs to only run the vehicle-detection algorithm during active calls, saving the phone’s battery.

Using a Nexus One phone, the researchers show that WalkSafe could reliably detect oncoming cars as far as 50 meters away from pedestrians (see the video above). They now plan to speed up the recognition algorithm to improve the app.

Now all we need is a system to alert people who text and walk.