11/08/2024
By Zubin Bhuyan
Ph.D. candidate name: Zubin Bhuyan
Defense Date: Friday, November 22nd, 2024
Time: 1 to 2 p.m. EST
Location: This will be a virtual defense via Zoom, Meeting ID: 943 7906 6888
Committee Members:
- Yu Cao, Ph.D. (Advisor), Professor, Director, Miner School of Computer & Information Sciences, UMass Center for Digital Health (CDH).
- Benyuan Liu, Ph.D. (Advisor), Professor, Director, Miner School of Computer & Information Sciences, UMass Center for Digital Health (CDH), Computer Networking Lab, CHORDS.
- Yuanchang Xie, Ph.D. (Co-advisor), Professor, Civil and Environmental Engineering.
- Hengyong Yu, Ph.D. (Member), FIEEE, FAAPM, Professor, Department of Electrical & Computer Engineering.
- Yan Luo, Ph.D. (Member), Professor, Department of Electrical & Computer Engineering.
Abstract:
Intelligent Transportation Systems (ITS) encompass a broad range of technologies aimed at enhancing the safety, efficiency, and sustainability of transportation and infrastructure. These systems manage traffic flow, bolster safety, and optimize public transit efficiency through innovations such as real-time tracking and scheduling. ITS rely on the continuous collection and intelligent analysis of data, enabling proactive management and informed decision-making in response to dynamic traffic conditions. Leveraging deep learning has led to significant advancements in traffic control, congestion reduction, and the development of autonomous driving technologies, thereby improving both system efficiency and safety. Despite these advancements, deep learning in ITS still holds substantial untapped potential, especially in managing complex environments and fulfilling the increasing safety and equity demands of contemporary transportation systems.
This doctoral thesis focuses on applying deep learning and computer vision (CV) techniques to a comprehensive set of datasets, including both ground-level and aerial imagery in RGB and thermal modalities. The primary goal is to enhance vehicle detection, tracking, and analysis, with a particular focus on highway traffic. Data collection spans multiple locations in Massachusetts and New Hampshire, addressing various general traffic settings and specifically targeted work zones to ensure the dataset’s applicability to real-world scenarios. Notably, three of these locations are active work zones, including two nighttime work zones with lane closures. These sites provide both aerial datasets, captured via drones, and ground-level thermal footage. The aerial datasets are annotated with Oriented Bounding Boxes (OBB), while segmentation masks are utilized in the ground-level thermal videos. We developed three key datasets: the MA Highway Aerial Dataset, the MA Nighttime UAV Thermal Dataset, and the MA-NH Highway Thermal Dataset, each including vehicle annotations as well as segmentation of roadways and various infrastructure elements like channelizer drums and cones. We implement a modified mosaic data augmentation strategy that utilizes segmentation-mask areas and targets "hard examples" based on missed detections in tracked trajectories.
A novel deep learning model was developed for segmenting and detecting vehicles using OBBs, along with an OBB tracking algorithm that extracts vehicle trajectories in aerial videos. Additionally, we introduced a CNN architecture, based on YOLOv8 enhanced with Group Shuffle Convolution modules, which significantly improved the detection and segmentation performance in thermal drone videos. Furthermore, we developed models to segment and analyze roadways and related infrastructures, such as highways, exit lanes, and channelizer drums. Leveraging these models enables accurate identification of vehicles near work zones and detection of those making last-minute lane changes in active work zones with lane closures. This strategy of simultaneously segmenting roadways and vehicles not only yields more detailed information about highway and work zone scenarios, but also enhances the predictive capabilities of our system. Such improvements can potentially increase the responsiveness of traffic management systems in real-time scenarios.
Furthermore, we introduce Selective-SHAI, an approach tailored for rapid and accurate inference aimed at detecting small objects, specifically focusing on critical regions within a video frame, such as highways. This method could markedly improve the responsiveness of traffic management systems in real-time scenarios, thereby contributing to the development of safer and more efficient transportation networks. This research not only advances the foundational theories of ITS but also proposes practical solutions poised to transform the management and monitoring of traffic environments.