Research On Video-based Lane Markings Detection Technology

With number of vehicles boosting,it is inconvenient for individuals to move.Their property and personal safety are threatened by traffic accident.The Intelligent Transportation System(ITS)have received extensive research attention since they can ensure convenient transportation and improve people’s quality of life.Intelligent vehicle detection algorithms,which are low cost,high speed and high accuracy,have been developed and cooperated with other detection means(such as laser radar,sensor detection)to achieve smart driving or intelligent traffic management.It is significant to study vehicle detection algorithms.Video-based vehicle detection may be easily affected by some interference from external circumstances such as the illumination change,shadows of trees and the other vehicles.To solve these problems,an improved pixel based adaptive segmenter(PBAS)is proposed to determine the Region of Interest(ROI).The proposed algorithm comprehensively improves the performance of the existing PBAS algorithm through several strategies,including reducing the decision threshold,updating the learning rate,reducing the computational complexity by gradient region partitioning,proposing the foreground cumulative counting rule and the “ghosting” elimination algorithm.Aiming at the problem of low detection rate and poor illumination robustness for a single Haar-like feature.The fusion characteristics of Haar-like and HOG are proposed,and the Adaboost+SVM classifier is used for classification,which reduces the number of classifiers and the number of features required for detection.Aiming at the problems of low detection rate and high computational complexity of basic Haar-like features and HOG features,a vehicle detection method based on clustering Haar-like feature and a vehicle detection method based on sparsity HOG feature are proposed.Through the test of road images under different scenes and several data sets,the experimental results demonstrate that our method can not only overcome the interference from external circumstances under complex road scene,but also detect vehicle with high speed,and it shows a good accuracy and robustness.