摘要
在恶劣天气下路面湿滑状况将发生显著变化,导致驾驶条件恶化,极易引发恶劣的交通事故.为了帮助交通管理部门实时获取路面湿滑状况信息,了解路面附着性能,并以此制定合理的车速控制和交通诱导策略,通过采集道路湿滑图像,利用设计的具体SVM分类器结合3种训练算法对道路湿滑图像进行分类学习训练,并通过训练后的SVM分类器对大量道路湿滑图像进行分类实验,对道路湿滑状态进行分类研究.实验结果表明:(1)寻找合适的核函数,选择合适的参数是使用SVM进行高效分类的一个重要因素;(2)在训练过程中,均方误差(MSE)能反映出分类器实现的正确率,SVM的训练本身的误差决定了分类的正确率,而且训练个数的增多带来了特征空间维数的增加,从而导致计算量的增大.
Wet and slippery conditions in inclement weather will be an extraordinary change and lead to worsening of driving conditions.It can lead to bad accidents easily.To help people of traffic management departments get the real-time road wet and slippery condition information and the road surface adhesion,they can formulate a reasonable speed control and traffic guidance strategies.By collecting wet and slippery roads images,we use SVM classifier combining three kinds＇ classification algorithm to do the wet and slippery road images classification of learning and training.And we use the SVM classifier by trained to classify the wet and slippery road images.The results show that：（1） it is an important factor of efficient use of SVM classification that found an appropriate kernel function and selected the appropriate parameters.（2） In the training,the Mean Square Error（MSE） reflects the correct classification rate of SVM implementation.Training error of SVM determine the correct classification rate.The number of training brought an increase of the feature space dimension,and lead to the increase of computation.
出处
《武汉理工大学学报（交通科学与工程版）》
2011年第4期784-787,792,共5页
Journal of Wuhan University of Technology（Transportation Science & Engineering）
基金
"十一五"国家科技支撑计划重点项目资助(批准号:2009BAG13A03)
