Back to results list
Show full item record
Please use this identifier to cite or link to this item:
|Title:||Automatic incident detection under no-rain and rain conditions||Authors:||Li, Xiangmin||Degree:||M.Phil.||Issue Date:||2014||Abstract:||This research presented in this thesis focuses on the development of automatic incident detection (AID) algorithms for use under no-rain and rain conditions. A new extended standard normal deviate (ESND) algorithm is proposed by extending the widely used standard normal deviate (SND) algorithm. The previous SND algorithm is modified with two extensions. In the first extension, the weighting method is adopted to enhance the reliability of detection results. In the second extension, the variation of input data within sampling periods is restricted to reduce false alarms for incident detection. The algorithm development is on the basis of the available data collected originally for journey time estimation purpose in Hong Kong. These data have been collected by both video traffic detectors and automatic vehicle identification readers. Instead of installing more expensive traffic detectors, the proposed ESND algorithm has proved feasible in effective traffic incident detection, with the use of the available data collected originally for journey time estimation purpose. In addition to the widely used traffic stream parameters such as traffic speed, flow and occupancy for incident detection, two new traffic stream parameters are proposed as incident indicators. They are (1) the coefficient of variation of speed at the upstream detector and (2) the correlation coefficient of speeds of two adjacent detectors. Traffic data collected from both single detector stations and dual detector stations are selected as the inputs for the proposed ESND algorithm. The proposed ESND algorithm is firstly extended to be a flow-dependent ESND algorithm for incident detection under no-rain conditions. The preincident traffic flow condition is considered explicitly in the detection logic to improve the detection performance under various traffic flow conditions. Historical traffic and incident data on a selected urban road section in Hong Kong are used for calibration and validation of the proposed flow-dependent ESND algorithm. Five existing AID algorithms are selected and calibrated for comparison with the proposed flow-dependent ESND algorithm with the use of the available data collected for journey time estimation purpose on the previously selected urban road section under no-rain conditions. The comparison results show that the proposed algorithm outperforms the five selected AID algorithms in terms of the detection rate, false alarm rate and mean time to detect.
The proposed ESND algorithm is then extended to be a more generalized flow-rain-dependent ESND algorithm for incident detection under both no-rain and rain conditions. The rain condition together with the preincident traffic condition are considered explicitly in detection threshold determination. Instead of the traffic flow, the volume/capacity ratio is adopted to indicate the preincident traffic condition. Compared to the discrete detection thresholds in the previous flow-dependent AID algorithms, continuous detection thresholds are adopted in the proposed flow-rain-dependent ESND algorithm. These continuous detection thresholds are generated by a generalized detection threshold function in which both preincident volume/capacity ratio and rainfall intensity are modeled. The proposed flow-rain-dependent ESND algorithm is applied to the urban road network under the Hong Kong Journey Time Indication System (JTIS) in order to examine the detection performance of the proposed algorithm on a territory-wide basis. In this research, the proposed flow-rain-dependent ESND algorithm is calibrated on the basis of available Hong Kong historical traffic, incident and rainfall intensity data collected on the urban road network under the JTIS. The detection performance of the proposed flow-rain-dependent ESND algorithm outperforms the flow-dependent ESND algorithm and the five selected AID algorithms on a territory-wide basis. It is shown in this research that the proposed flow-dependent ESND algorithm performed satisfactorily for incident detection in urban areas even when data are collected for journey time estimation purpose only. The proposed flow-rain-dependent ESND algorithm could be used for incident detection under both no-rain and rain conditions. This is of importance to cities with substantial rainfalls similar to Hong Kong.
|Subjects:||Traffic flow -- Mathematical models.
Traffic flow -- Data processing.
Hong Kong Polytechnic University -- Dissertations
|Pages:||xxi, 141 pages : illustrations ; 30 cm|
|Appears in Collections:||Thesis|
View full-text via https://theses.lib.polyu.edu.hk/handle/200/7527
Citations as of May 22, 2022
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.