作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (19): 195-198. doi: 10.3969/j.issn.1000-3428.2012.19.050

• 人工智能及识别技术 • 上一篇    下一篇

基于最优特征选择的车辆跟踪方法

彭丽荣,何育枫,刘文军   

  1. (昆明理工大学国土资源工程学院,昆明 650093)
  • 收稿日期:2011-11-29 出版日期:2012-10-05 发布日期:2012-09-29
  • 作者简介:彭丽荣(1987-),女,硕士研究生,主研方向:数字图像处理,人工智能,地理信息系统;何育枫、刘文军,硕士研究生

Vehicle Tracking Method Based on Optimal Feature Selection

PENG Li-rong, HE Yu-feng, LIU Wen-jun   

  1. (Faculty of Land Resource Engineering, Kunming University of Science and Technology, Kunming 650093, China)
  • Received:2011-11-29 Online:2012-10-05 Published:2012-09-29

摘要: 针对智能交通系统的车辆跟踪问题,提出基于最优特征选择的车辆跟踪方法。综合颜色、纹理和形状特征确定特征集合,采用线性鉴别分析方法从特征集合中选取最优特征,使用Mean Shift算法在最优特征下预测目标位置,根据目标匹配结果确定车辆的运行轨迹,利用特征平滑方法更新特征模型。实验结果表明,该方法适用于不同的公路监控场景,能够准确、有效地跟踪运动目标。

关键词: 车辆跟踪, 特征选取, Mean Shift算法, 特征模型更新, 密度概率, 方向梯度

Abstract: Aiming at vehicle tracking problem of intelligent transportation system, this paper proposes a vehicle tracking method based on optimal feature selection. The method integrates color, texture and shape characteristics to create a feature set, and applies Linear Discriminate Analysis(LDA) method to select optimal characteristics as input features of the Mean Shift algorithm, predicts of the target location, and determines the vehicle’s trajectory by matching targets. Feature model is updated by smooth method. Experimental results show that this method can track vehicle accurately and effectively in different highway monitor scenarios.

Key words: vehicle tracking, feature selection, Mean Shift algorithm, feature model update, density probability, direction gradient

中图分类号: