Research On Structured Lane Line Detection Based On Improved Deep Convolution Neural Network

Due to the change of modern transportation mode and the rapid development of automobile industry,automobile travel has become the main way of people’s daily travel.Because of the development of automobile technology at present,vehicles often travel at a very high speed.Drivers need to concentrate on the acquisition of road information when driving vehicles,and a little carelessness will cause serious consequences.In order to protect the safety of people’s lives and property,advanced auxiliary driving system is urgently needed.The detection of auxiliary driving technology for road environment is the key to realize the development of machine-led driving technology,and the lane line detection of structured road is the basis of the basic driving function of automobile with high performance auxiliary driving system.Compared with the traditional structured road lane detection algorithm,convolution neural network algorithm can effectively improve the robustness of the detection algorithm.Based on the deep convolution neural network,this paper studies the detection of structured road lane.The following research results have been obtained:(1)Based on the problems existing in the standard beetle whisker search algorithm,this section improves the standard beetle whisker search algorithm,designs the adaptive beetle whisker search algorithm(Automatic Beetle Antennae search algorithm,ABAS),the new algorithm retains the bit of the beetle whisker search algorithm,and optimizes the operation of fixed step size in the algorithm(2)In the field of computer vision,image segmentation is an important step in image processing and a prerequisite for subsequent computer image understanding and recognition.There are some defects in the traditional segmentation method.A new image segmentation algorithm is formed to solve the image segmentation problem,which is adaptive to the K-means mean clustering algorithm.the contrast experiment proves that the segmentation effect of this algorithm is better than that of Otsu algorithm,K-means mean clustering algorithm and particle swarm optimization algorithm.the new ABASK algorithm designed in this paper can preserve more image details while ensuring the segmentation effect.This chapter improves the feature extraction network Dark Net-53 of YOLOv3 algorithm,adjusts the structure of Dark Net-53 feature extraction network,designs a new feature extraction network Dark Net-25,and designs an improved YOLOv3 algorithm based on this feature extraction network.The convolution network structure of feature extraction is simplified,the number of parameters of convolution neural network is reduced,the operation quantity is improved YOLOv3 the depth convolution neural network algorithm balances the timeliness and accuracy of detection well.Compared with other YOLOv3 algorithms and YOLOv4 algorithms,structured road lane detection tasks can be better realized.