Research Of Key Technologies For Lane Departure Warning System Considering The Driver Characteristics

Lane departure warning system (LDWS), one of driver assistance system (DAS), canmake drivers easier and effectively decrease the crash accidents or mitigate the impact ofvehicle collisions, and thus is playing a crucial role to improve traffic safety. Traditionalmethod has poor performance on robustness of the driving environment recognition, theaccuracy of driver intention recognition and the accordance of pre-warning algorithm withdriver characteristics. So it has high false alarm rate, and is poor acceptance for drivers. Inorder to resolve above problems, after constructing the research platform, a LDWSconsidering driver characteristics is proposed, and the key technologies related to the systemare studied.In order to meet the experimental needs, simulation platform and real car test-bed forexperiments are designed. The hardware and software are presented in detail. Some testingexperiments are designed and implemented, after analyzing experimental results, a relevancemodel is proposed by utilizing normal and Gamma distribution. The model enables thedriving simulator to be used for the study of lateral driving behavior in replacement of theinstrumented vehicle in real traffic.According to the system structure, the key technologies related to the system areresearched. The results of image detection based on Adaboost algorithm and radar data areused to detect front vehicle by a fusion algorithm, which can promote the vehicle detectionaccuracy. Based on linear lane model, the curvature is estimated by the candidate points of theremote line fitting. Combined with the vehicle detection results, a lane detection method incrowded traffic conditions based on points set optimization is proposed which can decreasethe interference of the driving vehicles. Based on plenty of driver experiments in real road, anew lane departure waning algorithm considering the driver characteristics was proposed. Thedifferent thresholds of time to lane cross (TLC) were set when driver with different drivingbehaviors in different lanes and departure direction. In order to decrease false alarm rate, theconscious lane departure was distinguished from unconscious by real-time monitoring driverhead movement.To validate the effectiveness of the proposed method, a lane departure warning systembased on digital signal processor (DSP) with multi-sensor information fusion is developed onreal car test-bed. A LDWS test method is proposed. The simulation and real car test resultsshow that the environment recognition method has good robustness to various illumination,weather and road conditions, and the proposed LDW algorithm promote the adaptability tothe driver characteristics and decrease the false alarm rate, so it has favorable driveracceptance.