基于动态时间弯曲的多模板匹配车型分类

Multiple templates matching for vehicle’s classification based on dynamic time warping

中文关键词  动态时间弯曲  多模板匹配  车型分类  感应线圈  信息熵

英文关键词  DTW(dynamic time warping)  multiple templates matching  vehicle’s classification  inductive loop  information entropy

中文摘要

      感应线圈采集到的频率变化曲线, 取决于车速、车辆自身结构及其通过感应线圈时的相对位置. 针对同类车辆的频率变化曲线存在伸长或压缩情况, 研究一种基于动态时间弯曲(DTW)的车型分类新方法. 首先对采集到的车辆频率变化数据序列进行预处理; 其次, 结合DTW方法和最优聚类分析原理, 选取同类车型的多个频率变化特征曲线作为分类模板, 以改善单模板匹配方法的分类正确率; 再次, 确定同类车型的局部区分度指标, 并优化调整局部区分度指标范围; 最后对考察车辆的采样曲线进行多模板匹配分析, 得出车型分类结论. 仿真实验分析了本文方法与其他方法比较存在的优势.

英文摘要

      A vehicle’s frequency-response curve is determined by its velocity, structural framework and the relative location to an inductive coil detector. As the frequency-response curves of vehicles of the same type are varied in length, a novel classification method based on the dynamic-time-warping (DTW) algorithm is proposed. After a pretreatment on the obtained frequency-response curves, multiple curves are selected as vehicle classification-templates for DTW algorithm and the optimal clustering method. This will improve the classification properties of the single-template matching method. Next, we determine the dissimilarity-indices and their ranges for the vehicles of the same type. Finally, the classification result is obtained by matching the sample curve to the multiple templates using DTW algorithm. A simulation is given to illustrate the advantage of the method.