基于PHOG特征及支持向量机的弯道自动检测  

Automatic curve detection based on PHOG feature and support vector machines

作　　者：贾世杰[1] 郜瑞芹 

机构地区：[1]大连交通大学电气信息学院,辽宁大连116028

出　　处：《计算机工程与设计》2014年第7期2531-2535,共5页Computer Engineering and Design

基　　金：国家科技型中小企业技术创新基金项目(09C26222123243)

摘　　要：为了能准确地判断弯道情况,提出了一种基于机器学习的弯道自动检测方法。提取道路弯道训练图像的塔式梯度直方图(PHOG)特征,利用支持向量机对提取的特征进行训练形成分类模型;利用该模型和弯道测试图像的塔式梯度直方图特征对道路弯道情况进行预测。测试结果表明,该方法能够在理想天气和不同程度的恶劣天气下准确判断左弯道和右弯道,对于不同弯度的左右弯道,其平均分类准确率达90%以上。To accurately judge the curves,a method of curve automatic detection based on machine learning was put forward. Firstly PHOG feature of curve images including training and testing images was extracted.Secondly SVM was employed in trai-ning a classification model with the PHOG feature of training images,and then the classification model was used to predict the road curves with PHOG feature of testing images.This method could accurately determine the left and right curves,while its average classification accuracy of the left and right for different camber curves reached above 90%in the context of ideal weather and different degrees of bad weather.

关 键 词：弯道检测 机器学习 塔式梯度直方图 支持向量机 恶劣天气 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]