Vehicle Detection Based on the AND–OR Graph for Congested Traffic Conditions

In urban traffic video monitoring systems, traffic congestion is a common scene that causes vehicle occlusion and is a challenge for current vehicle detection methods. To solve the occlusion problem in congested traffic conditions, the authors have proposed an effective vehicle detection approach based on an AND-OR graph (AOG) in this paper. This method includes three steps: constructing an AOG for representing vehicle objects in the congested traffic condition; training parameters in the AOG; and, finally, detecting vehicles using bottom-up inference. In AOG construction, sophisticated vehicle feature selection avoids using the easily occluded vehicle components but takes highly visible components into account. The vehicles are well represented by these selected vehicle features in the presence of a congested condition with serious vehicle occlusion. Furthermore, a hierarchical decomposition of the vehicle representation is proposed during AOG construction to further reduce the impact of vehicle occlusion. After AOG construction, all parameters in the AOG are manually learned from the training images or set and further applied to the bottom-up vehicle inference. There are two innovations of this method, i.e., the usage of the AOG in vehicle detection under congested traffic conditions and the special vehicle feature selection for vehicle representation. To fully test the method, the authors have done a quantitative experiment under a variety of traffic conditions, a contrast experiment, and several experiments on congested conditions. The experimental results illustrate that the method can effectively deal with various vehicle poses, vehicle shapes, and time-of-day and weather conditions. In particular, the approach performs well in congested traffic conditions with serious vehicle occlusion.

• Record URL:
  • https://doi.org/10.1109/TITS.2013.2250501
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/oclc/41297384
• Supplemental Notes:
  • Abstract reprinted with permission of IEEE.
• Authors:
  • Li, Ye
  • Li, Bo
  • Tian, Bin
  • Yao, Qingming
• Publication Date: 2013-6

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; References; Tables;
• Pagination: pp 984-993
• Serial:
  • IEEE Transactions on Intelligent Transportation Systems
  • Volume: 14
  • Issue Number: 2
  • Publisher: Institute of Electrical and Electronics Engineers (IEEE)
  • ISSN: 1524-9050
  • Serial URL: http://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=6979

Subject/Index Terms

• TRT Terms: Automatic vehicle detection and identification systems; Machine vision; Methodology; Traffic congestion
• Uncontrolled Terms: Feature extraction; Occlusion
• Subject Areas: Data and Information Technology; Highways; I72: Traffic and Transport Planning;

Filing Info

• Accession Number: 01524717
• Record Type: Publication
• Files: TLIB, TRIS
• Created Date: May 1 2014 4:36PM