Freeway traffic state detection using CCTV stationary images.

详细信息   

• 作者：Yang ; Kaixin.
• 学历：Doctor
• 年：2012
• 导师：Ran,Bin,eadvisorRussell,Jeffreyecommittee memberNoyce,Davidecommittee memberWu,Chinecommittee memberQian,Zhiguangecommittee memberRan,Binecommittee member
• 毕业院校：The University of Wisconsin
• Department：Civil & Environmental Engr.
• ISBN：9781267620972
• CBH：3527677
• Country：USA
• 语种：English
• FileSize：3925927
• Pages：142

文摘

Traffic detection and data collection are essential to transportation planning and management. The CCTV Closed Circuit Television) traffic cameras have been increasingly deployed for traffic surveillance on major roadways by DOT Department of Transportation) in the U.S. However,the real-time traffic video data is still very limited to be published online because of the large load to the video servers and internet connections. As a result,the traffic snapshots which are clipped from the video data are widely used as an effective approach to publish the real-time traffic conditions through internet. Currently,almost all U.S. states can provide traffic snapshots via DOT or traveler information websites 511 System). These traffic snapshots are published and updated every 1 minute to 5 minutes in different states. They are potential new traffic data source that can be obtained free from internet without additional equipment and instruction cost. They can provide valuable traffic information to analyze traffic flow patterns and congestion status,where no other detectors are available. How to utilize these online CCTV snapshots effectively for traffic detection and data collection would be of practical significance. In this research,a novel traffic state detection model is developed,which could contribute to both academic and industry fields. The primary contribution of this study is to provide an effective solution to detect the traffic state using the CCTV stationary images. Traffic space occupancy is used as a reliable measure for quantitative traffic analysis and congestion estimation. The practical value of the proposed approach is that it turns widely freely available video source from internet into useful traffic information,and this could be done without interfering the existing CCTV system in DOT. The traffic image data used in this study were collected from I-894 freeway corridor in Milwaukee,Wisconsin. I proposed a model to measure the traffic space occupancy,which could not be measured directly in the field. In addition,a perspective transformation method is provided to estimate the traffic parameters without accurate camera calibration. The proposed algorithms and image processing were implemented in MATLAB. The model validation is performed with field loop detector data collected from the same roadway segments. The model performance was evaluated under different traffic status,illumination conditions,and roadway geometry. The validation and evaluation results indicate that the proposed model has a robust detecting capability for traffic status.