基于数字图像和支持向量机的交通路况检测  

Traffic conditions detection based on digital image and support vector machine

作　　者：翟玉婷[1] 贾世杰[1] 

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》2013年第12期4273-4277,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：为了避免物理检测方法对道路损坏严重、受天气因素影响大以及算法复杂等弊端,研究了利用支持向量机对不同程度交通路况图像进行分类的交通路况检测方法,采用塔式边缘方向梯度直方图作为图像的描述特征。实验结果表明,该方法能够提高恶劣天气条件下平均分类准确率,并且在各类天气情况下4类交通路况图像分类平均正确率达92%以上,能够有效检测不同程度交通路况。To avoid the drawbacks of the physical detection which damaged to the road seriously, affected by weather conditions and has complexity of the algorithm, a traffic conditions detection method using support vector machine to classify the different degrees of traffic conditions image is studied, using pyramid histogram of edge orientation gradients as the descriptive characteristics of the image. The experiments show that the studied methods could improve the average classification accuracy of the adverse weather conditions , and in all kinds of weather conditions, four categories of traffic conditions image＇s average classification accuracy is more than 92%, able to detect different degrees of traffic conditions.

关 键 词：交通路况检测 支持向量机 塔式边缘方向梯度直方图 恶劣天气 图像分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]