Knowledge Bank

Distracted driving is a common factor in many accidents in the United States. Car manufacturers are trying to implement technologies to increase a driver’s awareness of the road. While systems such as lane departure warnings, blind spot monitors, and adaptive cruise control are effective, there are few third party technologies and essentially none that are open source. The system described in this thesis uses a robust, cheap, and accurate vehicle detection system that is extendable, allowing for other parties to contribute to the system. The system detects key vehicle characteristics in a tiered process. Each tier prunes the set of hypothesized vehicle regions by requiring each region pass more refined characteristic criteria. Lane detection is also leveraged to reduce vehicle detection time. Testing shows that the system has a precision of 82% and a recall of 76%.