A Vehicle Forward Collision Warning System Combined With Curve Lane Detection

The vehicle Forward Collision Warning System(FCWS) is an important component of Advanced Driver Assistance Systems(ADAS). The FCWS can effectively reduce the collision accidents. Most of the existing vehicle FCWS based on radar is only applicable to straight lane, which cannot distinguish between vehicles ahead lie in its own lane and the ones lie in other lane without curve lane detection, hence lead to false warning and take a toll on the road safety. Because of this, this paper researches a vehicle forward collision warning system model by the method of computer vision, including curve lane detection method, accurate vehicle detection method and vehicle distance measurement method. This thesis is organized as follows:Firstly, in order to applied to challenging cases, robust curve lane detection based on Improved River Flow(IRF) algorithm and Random Sample Consensus(RANSAC) algorithm has been proposed in this paper. IRF algorithm can overcome the shortcoming of original River Flow(RF) algorithm against non-continuous lane. The feature points of curve lane on the far vision field can be searched by IRF algorithm, the model of hyperbolic-pair can be fitted by RANSAC combined with the linear parameters and curve lane can be detected robustly.Secondly, in view of the existing vehicle detection method cannot accurately detect the position of vehicle, especially vehicle tail bottom, an accurate vehicle detection method with the feature of the rear of the vehicle has been proposed in this paper. A region of interest about vehicle has been determined on the basis of rough vehicle detection. Then, the bottom rear of the vehicle and two sides of vehicle can be located accurately with the feature of the rear of the vehicle. In addition, the result of rough vehicle detection can be filtered by the proposed vehicle geometry constraints, which lead to lower the false detection rate.Finally, a fast and simple method of measuring the distance between vehicles by the model of pinhole imaging of camera has been proposed. According to the process of vehicle braking and the driver’s driving behavior, the safe distance of the vehicle can be determined. We design a vehicle forward collision warning system model combined with the key technologies. The demo of experiments shows that the proposed system can effectively distinguish between the vehicle on the own lane and the vehicle on the other lane, significantly avoid false warning.