计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (15): 251-258.DOI: 10.3778/j.issn.1002-8331.1905-0194

• 工程与应用 • 上一篇    下一篇

基于轨迹稀疏聚类的高速公路车辆检测

杨露,宋焕生,张朝阳   

  1. 长安大学 信息工程学院,西安 710064
  • 出版日期:2020-08-01 发布日期:2020-07-30

Highway Vehicle Detection Based on Sparse Trajectory Clustering

YANG Lu, SONG Huansheng, ZHANG Zhaoyang   

  1. School of Information Engineering, Chang’an University, Xi’an 710064, China
  • Online:2020-08-01 Published:2020-07-30

摘要:

针对高速公路中车辆的实时检测问题,提出了一种基于轨迹稀疏谱聚类的高速公路车辆检测方法。使用ORB算法检测特征点并利用基于金字塔LK光流算法进行跟踪得出特征点轨迹,将轨迹逆投影至三维世界坐标系,利用轨迹三维信息构建轨迹间的相似矩阵并对其进行稀疏化处理,采用谱聚类方法对特征点轨迹进行初步聚类,对谱聚类结果进行类间合并得出车辆检测结果。实验结果表明,方法花费了更少的时间代价,有效地解决了车辆遮挡问题,车辆实时检测精度提高至93%,具有一定的有效性和价值。

关键词: 车辆检测, 轨迹聚类, 谱聚类, 稀疏化

Abstract:

To solve the real-time vehicle detection problem in expressway, a vehicle detection method based on trajectory sparse spectral clustering is proposed. The main processing adopts ORB algorithm to detect feature points, tracks the trajectory of these feature points by Pyramid LK optical flow algorithm. The next step is inversely projecting trajectory to 3D world coordinate system. At the same time, similarity matrix is constructed with 3D trajectories information and is processed sparsely. After preliminary clustering the feature points, vehicle detection result is obtained by merging classes. Experimental results show that proposed method effectively solves vehicle occlusion problem and vehicle detection accuracy is improved to 93%, which has certain validity and value.

Key words: vehicle detection, trajectory clustering, spectral clustering, sparsity