摘要
为了适应复杂的车道线路况的识别,提出了应用Gist-SVM机器学习对直线型-弯曲型车道线自动检测分类的方法。首先通过Gist-SVM训练2种直线型和弯曲型分类模型;然后利用测试图像的特征与训练模型进行预测学习,应用支持向量机自动分类直线型和弯曲型车道类型;最后,检测的直线型车道线图像利用加约束Hough变换进行检测识别,检测的弯曲型车道线采用多数小线段直线拟合方法拟合弯道。同时设计一种适应本文所提方法的车道线检测识别系统的界面,将该车道线检测算法整合到该系统界面内。实验结果证明,采用Gist-SVM可自动检测分类车道线类型,该算法对直线型-弯曲型车道线检测识别的错检率减少20%,提高了检测的准确性。
In order to adapt to the complicated lane condition recognition,a method of automatically detecting and classifying linear-bending lane lines using Gist-SVM machine learning is proposed.Firstly,both linear and curved classification models are trained by Gist-SVM;then the characteristics of the test images and the training model are used for predictive learning,and support vector machines(SVMs)are used to automatically classify linear and curved lane types;and finally the detected linear lane lines are detected and identified using the constrained Hough transform,and the detected curved lane lines are fitted with a curved line using most of the small line segments.At the same time,an interface adapted to the lane detection and recognition system of this paper is designed,and the lane detection algorithm is integrated into the system interface.The experiment results prove that Gist-SVM can automatically detect the classification of lane lines,reduce the erroneous detection rate of linear-bent type lane detection by20%,and improve the accuracy of detection.
作者
魏玉东
杨先海
谭德荣
Wei Yudong;Yang Xianhai;Tan Derong(School of Mechanical Engineering,Shandong University of Technology,Zibo 255049)
出处
《高技术通讯》
EI
CAS
北大核心
2018年第9期867-873,共7页
Chinese High Technology Letters
基金
山东省自然科学基金(ZR2016EL19)资助项目
