Predictive brake warning at night using taillight characteristic

We present a technique used for frontal collision warning in an intelligent vehicle. The system uses an image processing algorithm to detect taillights (red brake lights) of vehicles driving at night time. The area of red lights, the size of the largest brake lights, and the vehicle's speed are used to compute the rear-end collision risk, which can be used as an indicator to danger that potentially leads to frontal collision. Therefore, the driver can be warned, and he can perform brake action on time. The algorithm uses a threshold to detect red color on an RGB image. The thresholds are obtained from statistical collection of 50 sample images. From our experiment on real driving situation, the collision risk can predict the actual brakes with 90% accuracy. The result also indicates that the area of red lights detected has a close relationship with the brake applied.