Yao Jan Wu

Video-based vehicle detection and tracking using spatio-temporal maps

Surveillance video cameras have been increasingly deployed along roadways over the past decade. Automatic traffic data collection through surveillance video cameras is highly desirable; however, sight-degrading factors and camera vibrations make it an extremely challenging task. In this paper, a computer-vision–based algorithm for vehicle detection and tracking is presented, implemented, and tested. This new algorithm consists of four steps: user initialization, spatiotemporal map generation, strand analysis, and vehicle tracking. It relies on a single, environment-insensitive cue that can be easily obtained and analyzed without camera calibration. The proposed algorithm was implemented in Microsoft Visual C++ using OpenCV and Boost C++ graph libraries. Six test video data sets, representing a variety of lighting, flow level, and camera vibration conditions, were used to evaluate the performance of the new algorithm. Experimental results showed that environmental factors do not significantly impact the detection accuracy of the algorithm. Vehicle count errors ranged from 8% to 19% in the tests, with an overall average detection accuracy of 86.6%. Considering that the test scenarios were chosen to be challenging, such test results are encouraging.

Climate Change Influence on Priority Setting for Transportation Infrastructure Assets

Transportation infrastructure could be vulnerable to local manifestations of global climate change, such as storm frequencies and durations of seasons. To adapt, transportation agencies need methodologies for reprioritizing their assets subject to the new sources of vulnerability. Prioritizing assets is nontrivial when criteria assessments and owner/operator preferences are considered in conjunction with the possible climate scenarios. Few efforts to date have addressed these scenarios in a priority setting for infrastructure asset management in the literature. This paper extends a scenario-based multicriteria decision framework that can assist decision makers in effectively allocating limited resources to adapt transportation assets to a changing climate. The framework is demonstrated with one of the most susceptible metropolitan transportation systems in the United States, the Hampton Roads region in coastal southeastern Virginia. First, the high-level goals of a long-range transportation plans are used in a traditional multicriteria analysis to generate a baseline prioritization of assets. Next, several scenarios that incorporate and combine a variety of climate conditions are identified. Finally, the scenarios are used to adjust the initial criteria weighting, which results in several reprioritizations of the assets. The results help to identify the most influential scenarios and characterize the sensitivity of the baseline prioritization across multiple scenarios. With these results, additional scientific and investigative efforts can be focused effectively to study and understand the influential scenarios.

DRIVE Net: E-Science Transportation Platform for Data Sharing, Visualization, Modeling, and Analysis

In past decades, transportation research has been driven by mathematical equations and has relied on scarce data. With increasing amounts of data being collected from intelligent transportation system sensors, data-driven or data-based research is expected to expand soon. Most online systems are designed to handle one type of data, such as from freeway or arterial sensors. Even if transportation data are ubiquitous, data usability is difficult to improve. A framework is proposed for a regionwide web-based transportation decision system that adopts digital roadway maps as the base and provides data layers for integrating multiple data sources (e.g., traffic sensor, incident, accident, and travel time). This system, called the Digital Roadway Interactive Visualization and Evaluation Network (DRIVE Net), provides a practical method for facilitating data retrieval and integration and enhances data usability. Moreover, DRIVE Net offers a platform for optimizing transportation decisions that also serves as an ideal tool for visualizing historical observations spatially and temporally. Not only can DRIVE Net be used as a practical tool for various transportation analyses, with the use of its online computation engine, DRIVE Net can also help evaluate the benefit of a specific transportation solution. In its current implementation, DRIVE Net demonstrates potential to be used soon as a standard tool to incorporate more data sets from different fields (e.g., health and household data) and offer a platform for real-time decision making.

A Google-Map-Based Arterial Traffic Information System

Web-based mapping technologies have been utilized for traffic information systems. However, most such systems are for freeways and very few of them focus on arterials or urban streets. This paper presents a real-time Google-map-based Arterial Traffic Information (GATI) system for urban streets in the City of Bellevue, Washington State. Open source web tools and emerging web technologies, such as Ajax, are used in implementing the system to ensure its performance and minimize its cost. Convenient administrative functions are enabled through advanced database design and the Model-View-Controller (MVC) application. This GATI system, though presented and demonstrated by using Bellevues data, is a general technology that can be applied to any arterial network.

Video-Based Monitoring of Pedestrian Movements at Signalized Intersections

Pedestrian and cyclist crossing characteristics are important for the design of urban intersections and signalized crossings. Parameters such as waiting time, crossing time, and arrival rate are key variables for describing pedestrian characteristics and improving crossing designs and signal timing plans. Manually collecting such data is often extremely labor intensive. Therefore, an automated computer-vision-based approach is introduced for collecting these parameters in real time with ordinary video cameras. Broadly defined pedestrian objects, including bicyclists and other nonmotorized modes, are extracted by means of the background subtraction technique and tracked through an inherent cost characteristic function in conjunction with an α-β-filter. The waiting-zone concept introduced helps provide robust pedestrian tracking initialization and parameter extraction. The proposed approach is implemented in a pedestrian tracking (PedTrack) system by using Microsoft Visual C++. Tested with real video data from three study sites, this system was proved to be effective and about 80% of pedestrian crossing events were successfully detected. PedTrack shows the potential to be a great data collection tool for nonmotorized object movements at intersections.

Traffic Monitoring and Vehicle Tracking using Roadside Cameras

This paper studies the integration and implementation of digital image processing techniques on the roadside camera for traffic monitoring and vehicle tracking. The image processing framework developed in this study is mainly composed of five stages: (1) pre-processing, (2) foreground segmentation, (3) shadow removal, (4) tracking, and (5) traffic parameters extraction. During the pre-processing stage, the information of road geometry is obtained and the camera is calibrated. At the foreground segmentation stage and shadow removal stage, moving vehicles are segmented from the original input images. To make the system more robust, an alpha-beta filter is used at the multi-vehicle tracking stage. Subsequently, related traffic parameters are extracted at the end of each tracking mechanism. The experimental results show that this system is capable of successfully extracting the traffic parameters, including the trajectory of the moving vehicles based on the image sequences captured by a digital camera on a free flow traffic in the daytime..

Modeling animal-vehicle collisions using diagonal inflated bivariate Poisson regression.

Two types of animal-vehicle collision (AVC) data are commonly adopted for AVC-related risk analysis research: reported AVC data and carcass removal data. One issue with these two data sets is that they were found to have significant discrepancies by previous studies. In order to model these two types of data together and provide a better understanding of highway AVCs, this study adopts a diagonal inflated bivariate Poisson regression method, an inflated version of bivariate Poisson regression model, to fit the reported AVC and carcass removal data sets collected in Washington State during 2002-2006. The diagonal inflated bivariate Poisson model not only can model paired data with correlation, but also handle under- or over-dispersed data sets as well. Compared with three other types of models, double Poisson, bivariate Poisson, and zero-inflated double Poisson, the diagonal inflated bivariate Poisson model demonstrates its capability of fitting two data sets with remarkable overlapping portions resulting from the same stochastic process. Therefore, the diagonal inflated bivariate Poisson model provides researchers a new approach to investigating AVCs from a different perspective involving the three distribution parameters (λ(1), λ(2) and λ(3)). The modeling results show the impacts of traffic elements, geometric design and geographic characteristics on the occurrences of both reported AVC and carcass removal data. It is found that the increase of some associated factors, such as speed limit, annual average daily traffic, and shoulder width, will increase the numbers of reported AVCs and carcass removals. Conversely, the presence of some geometric factors, such as rolling and mountainous terrain, will decrease the number of reported AVCs.

A vision-based vehicle behavior monitoring and warning system

This paper outlines the approach to restrain the transportation (or car) accidents caused by dangerous vehicle behaviors, especially those of lane departure and speeding. Based on the proposed approach, the possibility of car collision is reduced by an equipped in-vehicle vision-based system that monitors the sight in front of the car and issues certain necessary warning. Meanwhile, the infrastructure of monitoring and warning system as well as related image processing techniques is proposed. Furthermore, an application for decision model to launch warning is also discussed herein. Finally, the proposed approach is validated in real road tests.

Measuring Signalized Intersection Performance in Real-Time With Traffic Sensors

Safety and quality of travel on arterial networks tie closely to the performance of signalized intersections. Measures commonly used for intersection performance evaluations are control delay, queue length, and cycle failure. However, these variables are not directly available from typical configurations of traffic sensors designed for intersection signal control. Collecting vehicle control delay data manually for intersection performance measurement has been a task too time-consuming and labor-intensive to be practical. Video image processors (VIPs) have been increasingly deployed for intersection signal control in recent years. This study aims to use the extra detection capabilities of VIPs for performance monitoring at signalized intersections. Most VIPs can support up to 24 virtual loops, but normally less than half of the virtual loops are used. By properly configuring the spare virtual loops and analyzing the loop measurements, intersection performance can be monitored in real time. In this research, we propose an approach for measuring queue length and vehicle control delay at signalized intersections based on traffic count data collected with traffic sensors. This algorithm has been implemented in a computerized system called In-PerforM. The In-PerforM system was evaluated by both field tests and simulation experiments. Although the VIPs’ counting errors do affect the accuracy of field test results, we still received encouraging results on queue lengths and control delay measurements in both the field tests and simulation experiments. This demonstrates that the In-PerforM system, and therefore the proposed algorithm, has the potential to be a cost-effective approach for performance measurement at signalized intersections.

Climate Change Effects on Transportation Infrastructure

The potential effects of climate change on transportation infrastructure have been receiving attention in recent years. An especially useful and increasingly common approach to investigating the potential effects of climate change on infrastructure is the use of geographic information systems (GISs) for risk analysis because climate change effects are likely to occur in conjunction with other geographically specific impacts such as storm surge and traffic operations, whose vulnerability can be most effectively quantified with GIS-based tools. To demonstrate the efficacy of these tools, a scenario-based risk analysis approach is presented: it investigates the effects of climate change on transportation infrastructure in Hampton Roads, Virginia. First, climate change effects in the study site are investigated to develop representative climate change scenarios. Then, a GIS-based evaluation of transportation infrastructure vulnerability to sea level rise and storm surge is formed by combining the GIS data set with results from the Sea, Lake, and Overland Surges from Hurricanes (SLOSH) model. Finally, the proposed risk model generates a GIS-based risk map under three scenarios of climate change threat. Results indicate that the city of Virginia Beach, Virginia, is at high risk in all three scenarios because of climate change events, a high level of transportation activity, and density of transportation facilities. The risk map—a visualization of the risk model—can assist transportation planners and decision makers with prioritizing assets to allocate resources for emergency preparation and response.