The Design And Implementation Of A Lane Departure And Anti-collision Warning System

With the improvement of people's quality of life,the requirements for cars are not limited to being able to engage in transportation between the two places.In addition,it is necessary to ensure the driving safety and reduce the occurrence of traffic accidents.Lane departure warning and collision warning systems as an important part of current vehicle safety technology.It needs to estimate the driving safety,according to the surrounding environment.Remind the driver by sound,vibration or other forms,and prompt them to take corresponding actions.This article mainly studies the lane line deviation and collision warning system,the content is as follows:Aiming at the problem of low recognition rate in low light environment.This paper propose an improved Retinex image enhancement algorithm to achieve adaptive correction.In this algorithm,the image brightness and contrast are modified by gain-offset operator based on global illumination.Experiments show that the algorithm can be applied to image enhancement and noise suppression under different illuminance conditions.The algorithm has nearly 25% faster than traditional Retinex image enhancement algorithm.In the process of detecting the lane edge,the noise in the image is close to gradient amplitude of the edge of low-intensity lane,resulting in the loss of lane line features.Combined morphological gradient information and gradient direction information as the screening condition for Canny edge detection.According to the current position of the vehicle and the lane line model,the distance between the vehicle and the lane line is set to realize the warning of vehicle departure.Experiments show that the method has over90% accuracy under different illuminance conditions.In the collision warning system,in order to reduce the overall cost of the vehicle detection system.In this paper,we combined the target detection network YOLO-V3 and the lightweight feature extraction network Mobile Net as the vehicle detection network.The combination of depthwise and pointwise convolution is used to replace the conventional convolution,and it can effectively reduce the amount of network computation and the number of model parameters.Finally,combined the vehicleposition information with the longitudinal road model,a multi-stage collision prevention and warning strategy is designed.Compared with the original network,this algorithm improves the detection speed by nearly twice..