基于支持向量机的车辆行为分析方法研究

Driving behavior analysis based on support vector machines for visual traffic surveillance

随着我国经济的快速发展,汽车保有量以年均8%的速度增加,给人民出行带来便利的同时也带来了诸多的交通问题:交通事故频发,交通运载效率降低。统计结果表明这些问题主要是由驾驶员错误或违规的交通行为导致的。针对这些问题提出一套能够实时地对行驶车辆的驾驶行为进行监控和分析的方法,该方法由车辆检测、车辆跟踪、车辆行为分析3部分组成。研究中分别对每部分的算法进行了筛选和部分改进,手工绘制了车辆典型驾驶行为的行驶轨迹样本,并采用支持向量机(SVM)的机器学习方法进行了训练。实验表明该方法能够准确地对被监控车辆的行驶行为进行判断,从而实现对违规车辆的驾驶行为进行有效地监控。

With the rapid development of our economy, the number of automobiles is growing for 8% in every year , which brings people＇ s convenience , at the same time, which brings a lot of traffic problems. Traffic accident happened frequently, and traffic carry capacity became insufficient. Statistics result showed that one main reason was caused by driving behavior mistakes. In this paper, we will present a real-time method that can detect, track and analyze the driving vehicles＇ behaviors. The proposed method contains three parts, which are vehicle detection, vehicle tracking and driving behavior analysis. Each part＇ s algorithms were filtered and were improved separately , and I got the typical driving behavior sam- ple through hand-drawing vehicle＇ s travel path, and use Support Vector Machine （SVM） for machine learning Experimental results showed that it can accurately judge monitored vehicle＇ s driving behavior, and realize efficient surveillance for vehicles＇ the driving behavior who was in violation.