Lane Line Detection Based On Edge Feature Fusion And Deep Neural Network

With the improvement of people’s living standard,cars have entered people’s life.Subsequently,the increasing traffic safety problems aroused people’s attention.Lane line detection,as an important part of automatic driving and auxiliary driving system,has been studied a lot.Its core idea is to detect the position of lane line accurately and quickly.Traditional lane line detection method to detect the image edge detection,threshold processing and curve fitting,by extracting the characteristics of the lane line detection lane line,rely on manual operation,the algorithm complexity is big,the lane line detection efficiency is low,and based on the model of the system is hard to modeling was carried out on the road scene change,so the traditional methods often produce robustness problems.In order to solve the problems existing in traditional lane line detection,this paper proposes a lane line detection scheme based on deep neural network.The proposed encoder-decoder convolutional neural network model is used to automatically extract the effective lane line feature information from the image to be detected,which can realize the end-to-end detection of the lane line image.In addition,considering the difference between positive and negative samples in lane line detection,we modify the loss function.Experiments show that the algorithm is simple and fast,and can effectively solve the problem of large difference between positive and negative samples.Considering that the lane line has obvious edge characteristics,this paper proposes a lane line detection scheme based on multi-channel feature fusion and convolution neural network by referring to the steps of edge detection in traditional schemes.Multichannel fusion technology is adopted to fuse the original image information with the lane line edge feature information,combine the multi-channel feature with deep learning,and use neural network training to obtain the information that effectively represents the lane line feature.The experimental results show that the detection effect based on edge feature fusion technology is better than the original RGB image input.The experiment also proves that the proposed network model has strong robustness and generalization ability.