Improvement of Forward Collision Warning in Real Driving Environment Using Machine Vision

摘要

Most forwarding collision warning systems utilize a kinematic model of vehicle’s deceleration, which uses only velocity and distance to the front cars. In our work, we demonstrated that by adding driving environment data extracted from machine vision to the model, the performance of the warning system can be significantly improved. We compared the performance of our forward collision warning system with and without the data from machine vision on real driving experiments on city roads. The results showed that the machine vision techniques can increase the accuracy of warning from an average of 53% to 73% when compare the warning signals from the system to the actual brakes.