Technology Research On Reliability Enhancement Of Automatic Incident Detection For Freeway Based On Fixed Vehicle Detector Data

Automatic Incident Detection(AID) is an important support tool for highway operation and management, and its detection results depends on a large number of real-time data obtained from the sensor, in practice, due to the quality of t he data is not reliable event detection systems often lead to unforeseen false alarms, repeating alarms and leak alarms, etc., which directly affects the credibility of the system work. C urrently, data reliability and engineering academic research on freeway incident detection is still relatively scarce which can not meet the actual application requirements. To this end, it has important practical significance to take a in-depth study of a variety of data unreliable factors on the freeway AID system, and then explore ways to enhance the detection results which will improve the reliability of AID results and provide a more reliable perception in freeway incident management to promote highway operation and management level.Aiming above freeway incident detection data reliability issues with fixed vehicle detector data as a typical case, this paper started from analyzing the impact to AID, and discover events false alarms, repeating alarms and leak alarm and other causes, and then study the reliability upgrade approach of AID. The main work is as follows:â‘  Impact analysis of unreliable data. First, through the analysis of typical data unreliable factors in the AID found to affect the detection results mainly from missing data, data anomalies and data is not synchronized three areas. Then, for the situation of missing data, establish a integrity assessment method of data in the perspective of AID. For the abnormal data, a correctness assessment method is proposed based on statistical and traffic flow theory, for data synchronization problem, with traffic simulation software and GISDK Transmodeler development tools, the mechanism of data sync is charactered qualitatively and quantitatively. Finally, the classic California algorithm, for example, analyzed three factors event data is not reliable detection of false positives, and varying degrees of impact reported repeated omissions, and lay the foundation for future improvements.â‘¡ Respectively, from the level of data preprocessing and algorithms proposed methods to enhance system reliability. Against data missing and data anomalies, made a series of data screening and repair mechanisms. For the data synchronization problem, by using the relative drift time, proposed a improved California algorithms. The practical application of proven, in the presence of data is not synchronized, the improved algorithm compared with the classical algorithm, the detection rate increased by 6.25 percent, 15.24 percent lower rate of false positives, false negative rate decreased 6.25%.Finally, the above data preprocessing methods and improved algorithm using VS2010 software implementation and practical application of C hongqing freeway Automatic Incident Detection, the results show that: Compared with the classical algorithm California, 6.66% detection rate, false alarm rate decreased 21.54 % false negative rate decreased 6.66%. Therefore, the results of this study can effectively reduce the negative impact of the data is not reliable factors on freeway Automatic Incident Detection to improve the credibility and reliability of the freeway incident detection results.