Home
Resource
Longitudinal scanline based spatial-temporal modeling and processing for vehicle trajectory reconstruction
PDF
PDF format is widely accepted and good for printing.
Plug-in required
PDF-1(5.85 MB)
Citation & Export
View Usage Statistics
Staff View

Citation & Export
Hide

Simple citation

Zhang, Tianya. Longitudinal scanline based spatial-temporal modeling and processing for vehicle trajectory reconstruction. Retrieved from https://doi.org/doi:10.7282/t3-n71q-8g72

Export

Click here for information about Citation Management Tools at Rutgers.

Statistics
Hide

Description

TitleLongitudinal scanline based spatial-temporal modeling and processing for vehicle trajectory reconstruction
NameZhang, Tianya (author); Jin, Peter (chair); Gong, Jie (member); Liu, Xiang (member); Piccoli, Benedetto (member); Rutgers University; School of Graduate Studies
Date Created2022
Other Date2022-05 (degree)
SubjectTransportation, Computer science, Artificial intelligence, Intelligent transportation system, Roadside camera, Spatial temporal map, Vehicle trajectory
Extent165 pages : illustrations
DescriptionThe Artificial Intelligent (AI) and Connected Automated Vehicle (CAV) technology allow innovative approaches to gather and analyze traffic data, helping traffic managers to proactively adjust the operational plan through real-time monitoring and analysis. Among many traffic data collection methods, video sensors hold high promises, providing extensive coverage and containing complete vehicle dynamic information. With the correct data, Intelligent Transportation System (ITS) techniques will be more beneficial through better understanding and optimizing traffic, simultaneously improving traffic safety while reducing congestion and emissions. This research explores longitudinal scanline-based traffic video analytics, ranging from intermediate-angle to high-angle cameras. Scanline-based image processing for spatial-temporal modeling has been thoroughly investigated and examined using different levels of measurements of effectiveness (MOE). The method of coordinate transformation is through three systems (STMap Coordinate, Video Coordinate, World Coordinate). Considering the current trend of 3D infrastructure modeling, the LiDAR model is used to assist in the fast calibration of a traffic camera. Two novel neural networks were designed in this dissertation, including Res-UNet+ and Dual Attention Inception Neural Networks (DAIU). Deep-learning models for STMap-based vehicle detection overcome several issues in the image processing method, such as static noises, ineffective background modeling, low-quality video data. The proposed deep learning models outperform several baseline models with superior detection performance and fast training speed. In contrast to other models that require frame-by-frame labeled training data, the longitudinal scanline significantly reduces the data annotation load by labeling only the trajectories in STMap, which comprises only a small fraction of the overall video. Furthermore, the STMap-based vehicle detection method is a one-stage solution rather than the two-stage tracking-by-detection paradigm. With STMap method, the vehicle detection and tracking problem is converted into image segmentation efforts. The model comparison results with the state-of-the-art vehicle tracking solution show that the proposed method can achieve better overall performance and be more robust in challenging scenarios. In the connected and automated transportation system era, the camera-based traffic analytic technique has become indispensable. With data-driven technology, we can connect and empower traffic engineers, city agencies, and planners to build smart communities by mitigating traffic congestion and making the road safer for vehicles, cyclists, and pedestrians.
NotePh.D.
NoteIncludes bibliographical references
Genretheses
Persistent URLhttps://doi.org/doi:10.7282/t3-n71q-8g72
LanguageEnglish
CollectionSchool of Graduate Studies Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Version 8.5.5
Rutgers University Libraries - Copyright ©2024