Lane Departure Warning Algorithm Based On Fused Image And Lidar Data

In this paper,the lane departure warning system with fused image and lidar data is studied.Lane departure warning system is one of the key technologies in the advanced driver assistance systems.It is mainly used to prevent the vehicle from deviating from the current driving lane and causing traffic accidents.This paper first analyzes the components of the lane departure warning system,and summarizes the three major difficulties of the current study,and for each difficulty to carry out relevant research.The main research problems and achievements are summarized as follows:1.Lane detection under complicated road condition.The current lane detection methods has achieved good results under stable lighting conditions and good road conditions.The difficulty lies in the detection of lane under complex conditions,such as strong changes in light,shading,obstruction,road wearness,dashed lane line and other complex road conditions.In this paper,we propose a lane detection algorithm for complex road conditions.Feature extraction is achieved in gray scale space and HSV space to reduce feature loss.Aiming at the problems of obstacle obstruction,shadowing and other road marking,the influence of obstruction on lane detection is reduced by using the method of gradient direction judgment which utilize the parallel relationship between two edges of one lane line.Aiming at the lane wearness and dashed line detection,the improved river flow algorithm is used to connect the features.2.The high false alarm rate and the unstable warning information of the lane departure warning model.The false alarm will interfere with the driver's attention,reduce the system availability,and even erroneously change the vehicle motion,causing danger.The instability of the warning information can cause the driver unable to judge whether the current driving state is dangerous and whether to correct the vehicle trajectory.In this paper,the advantages and disadvantages of the existing departure warning model are analyzed,and an staged departure warning model is proposed.It can reduce the false alarm rate and improve the stability of the warning information,while maintaining the detection rate and warning time.3.The low reliability of lane detection with single sensorMost current lane departure warning studies use visual sensors as the sole source of road data,which is less reliable in the actual usage.On the one hand,when there is a sensor error,no other measurement results are used for comparison and correction.On the other hand,it can't adapt to the switching of structured road and unstructured road.In this paper,a method of fusion of lidar and image data at feature level is proposed.Firstly,a parallel line model is proposed based on the actual road structure for two kinds of sensor measurement data.Secondly,a method of data fusion is proposed to reduce the influence of single sensor random errors on the system detection results based on prior information and historical data.