Back to results list
Please use this identifier to cite or link to this item:
|Title:||Self-adaptive vision-based vehicle reidentification system for dynamic freeway travel time estimation||Authors:||Wang, Jiankai||Keywords:||Travel time (Traffic engineering) -- Measurement.
Motor vehicles -- Automatic location systems.
Hong Kong Polytechnic University -- Dissertations
|Issue Date:||2014||Publisher:||The Hong Kong Polytechnic University||Abstract:||Due to its potential for anonymous vehicle tracking, vehicle reidentification (VRI) has emerged as an effective approach for directly estimating freeway travel time. Asreal-time video of traffic scenes contains a wealth of vehicle information that may not be available from conventional detectors (e.g. inductive loops), there is a growing interest in development of vision-based VRI system. To further improve the robustness of VRI system against the potential changes in traffic condition, a self-adaptive time window component is also required. To this end, the thesis aims to contribute to the development of a self-adaptive vision-based vehicle reidentification (VRI) system for dynamic freeway travel time estimation. As a preliminary investigation, the thesis first considers developing a basic vision-based VRI system under static traffic conditions. Various vehicle feature data (e.g. color,length and type) are extracted from the video record, and a data fusion approach is then introduced to combine these features to generate a probabilistic measure for reidentification decision. The vehicle-matching problem is then formulated as a combinatorial problem and solved by a minimum-weight bipartite matching method. The proposed basic vision-based VRI system is then extended and applied for automatic incident detection under free condition. The relatively high matching accuracy of basic VRI would allow for a prompt detection of the incident vehicle and, hence, reduce the incident detection time. An enhanced vehicle feature matching technique is adopted in the basic VRI component to explicitly calculate the matching probabilities for each pair of vehicles. Also,a screening method, which is based on the ratios of the matching probabilities, is introduced to reduce the false alarm rate. The basic VRI is also extended to the static case where multiple video cameras exist. A hierarchical Bayesian matching model is then proposed to consider vehicle reidentification over multiple detectors as an integrated process such that the overall matching accuracy could be improved. For the dynamic traffic conditions, the thesis introduces an additional self-adaptive time window component into the basic VRI system to improve its performance in terms of dynamic travel time estimation. Specifically, an iterative VRI system with temporal adaptive time window constraint is proposed. To capture the traffic dynamics in real-time, the inter-period/temporal adjusting based on exponential smoothing technique is introduced to define an appropriate time window constraint for the basic VRI. To handle the non-predictable traffic congestions, the modified basic VRI is performed iteratively (i.e. iterative VRI) such that the improved VRI is capable of adjusting its parameters automatically. Finally, the thesis focuses on developing an integrated self-adaptive VRI system for a freeway with multiple video cameras under dynamic traffic conditions. The spatial dependencies between the travel time over different freeway segments are utilized for the further adjustment of the time window constraint. An iterative VRI system with spatial-temporal adaptive time window constraint is then proposed to cope with the purpose of dynamic travel time estimation.||Description:||xvi, 199 leaves : illustrations ; 30 cm
PolyU Library Call No.: [THS] LG51 .H577P CEE 2014 WangJ
|URI:||http://hdl.handle.net/10397/7180||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b27575391_link.htm||For PolyU Users||203 B||HTML||View/Open|
|b27575391_ir.pdf||For All Users (Non-printable)||7.75 MB||Adobe PDF||View/Open|
Citations as of Oct 14, 2018
Citations as of Oct 14, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.