Electronic Rearview Mirror Image Transmission And Nighttime Lane Line Detection Method

While cars bring great convenience and economic benefits to our life and production,the personal injury and economic loss caused by traffic accidents can not be ignored.Therefore,the number of traffic accidents can be reduced through the intelligent driving assistance technology of vehicles.Lane line detection is one of the key technologies of vehicle intelligent driving assistance.Environment during the day to detect lane line technology is relatively mature,but in the night environment,overall image brightness can cause some dark,light concentration distribution area is too bright and other areas have dark cause lane line difficult to extract the characteristics of the environment using a good algorithm is difficult to apply in the daytime,most of the current mainstream lane line detection technology using the neural network method and dependent on highperformance graphics processors,but the car can’t placed high performance graphics processor,and part of the lane line detection algorithm calculated,using embedded processors but embedded processor can’t meet the real-time requirements.Therefore,this paper studies the lane line detection in the nighttime environment,and the main work is as follows:1.In view of the situation that the concentrated distribution of light in the night environment will cause too bright in some areas and too dark in other areas,which will make it difficult to distinguish the lane line from the road,An adaptive gamma transform image enhancement method based on multi-scale Gaussian function is proposed.Using the global average gray value of the original image in the night environment as the reference value,three values are selected as the scale factors in the Gaussian function to generate the Gaussian template,and then the light component is obtained by convolution with the original image in the night environment.Then the gamma correction function is established according to the distribution characteristics of illumination components in the whole image.The experimental results show that this method can better increase the contrast between lane and road.2.In view of the traditional Canny operator edge detection noise exist in the process of making the effect is not ideal and could not adaptive double thresholds are adjusted to filter the edge problem,put forward based on the nonlocal average filtering of adaptive Canny operator linear edge extraction,using nonlocal average filtering method to remove the image noise,compared with the traditional Canny operator better retain the image details of gaussian filter does not cause the image blur,secondly use Otsu algorithm to calculate the double threshold of low threshold,high and low threshold value based on Canny operator recommended a ratio of 2:1 principle to calculate the high threshold,Experimental results show that the improved Canny edge extraction algorithm is more accurate to extract the night lane edge.3.For the car can not place high performance graphics processor and embedded processor computing power is insufficient,This paper uses the network to send the image data collected by the vehicle terminal to the remote server,and transmission using TCP protocol communication will produce glue bag problem,this paper puts forward a solution to stick package problems in the process of network transmission method,after analysis of package problem to design the lane line image resolution at night send combination with receiving solution,this will make every frame will complete to send with the next frame completely separate,and this method is verified by experiment with good reliability.