摘要
在以三轴数字式AMR(各向异性磁阻)传感器为核心设计车辆检测系统的基础上,研究不同车型的车辆对三维地磁场的扰动特性,设计了一种车型实时分类算法。从整车长度、车型尺寸、传动轴数目这3个视觉上直观反映车辆类型的属性出发,抽象出相对车长、平均能量、归一化方波的波峰/波谷数目3个波形特征量作为依据,根据分层决策树算法将车辆分为5大类。研究结果表明:算法研究地磁场的三维扰动信息,使分类结果一定程度上不依赖于行车速度和传感器的位置,从停车收费车道和不停车收费车道2种情况下的对比情况来看,算法确实消除了车速对车型分类的影响,能够有效地实现车型的实时分类。经过现场试验测试,分类准确率基本上达到80%以上,尤其对重型挂车的分辨率甚至可达到95%。
Abstract： This paper studied the nature of geomagnetic disturbance which was induced by differ- ent vehicles. A real-time classification algorithm was designed on the basis of the vehicle detec- ting system which took the 3-axis digital compass IC HMC5843 as core. As the vehicles＇ length, size and axle number are all closely related to vehicle types directly on the vision, the correspond- ing characteristics of collected waveforms of relative length, average energy and normalized square wave peaks were abstracted as the standards of classification. Vehicles were classified into 5 types by the decision tree. The results show that this classification does not depend on the vehi- cle speed and placement location of the sensors as the algorithm taking the three-dimensional dis- turbance information for research . Compared the data from ETC lanes with artificial tolling lanes, the influence of vehicle speed is eliminated indeed by the classification. The system is able to automatically detect vehicles, extract features, and classify them. The classification rate for heavy trunks can reach 95% and for other four types, it can reach about 80% by actual tests. 1 tab, 6 figs, 10 refs.
出处
《长安大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2013年第5期118-123,共6页
Journal of Chang’an University(Natural Science Edition)
基金
交通运输部科技项目(2010 319 812 080
2011 318 812 260)
关键词
汽车工程
车辆检测系统
车型分类
有限状态机
分层决策树
automobile engineering
vehicle detecting system
vehicle classification
finite statemachine
decision tree
