Benjamin Coifman

A REAL-TIME COMPUTER VISION SYSTEM FOR VEHICLE TRACKING AND TRAFFIC SURVEILLANCE

Abstract Increasing congestion on freeways and problems associated with existing detectors have spawned an interest in new vehicle detection technologies such as video image processing. Existing commercial image processing systems work well in free-flowing traffic, but the systems have difficulties with congestion, shadows and lighting transitions. These problems stem from vehicles partially occluding one another and the fact that vehicles appear differently under various lighting conditions. We are developing a feature-based tracking system for detecting vehicles under these challenging conditions. Instead of tracking entire vehicles, vehicle features are tracked to make the system robust to partial occlusion. The system is fully functional under changing lighting conditions because the most salient features at the given moment are tracked. After the features exit the tracking region, they are grouped into discrete vehicles using a common motion constraint. The groups represent individual vehicle trajectories which can be used to measure traditional traffic parameters as well as new metrics suitable for improved automated surveillance. This paper describes the issues associated with feature based tracking, presents the real-time implementation of a prototype system, and the performance of the system on a large data set. ©

A real-time computer vision system for measuring traffic parameters

For the problem of tracking vehicles on freeways using machine vision, existing systems work well in free-flowing traffic. Traffic engineers, however, are more interested in monitoring freeways when there is congestion, and current systems break down for congested traffic due to the problem of partial occlusion. We are developing a feature-based tracking approach for the task of tracking vehicles under congestion. Instead of tracking entire vehicles, vehicle sub-features are tracked to make the system robust to partial occlusion. In order to group together sub-features that come from the same vehicle, the constraint of common motion is used. In this paper we describe the system, a real-time implementation using a network of DSP chips, and experiments of the system on approximately 44 lane hours of video data.

Estimating travel times and vehicle trajectories on freeways using dual loop detectors

Recent research has investigated various means of measuring link travel times on freeways. This search has been motivated in part by the fact that travel time is considered to be more informative to users than local velocity measurements at a detector station. But direct travel time measurement requires the correlation of vehicle observations at multiple locations, which in turn requires new communications infrastructure and/or new detector hardware. This paper presents a method for estimating link travel time using data from an individual dual loop detector, without requiring any new hardware. The estimation technique exploits basic traffic flow theory to extrapolate local conditions to an extended link. In the process of estimating travel times, the algorithm also estimates vehicle trajectories. The work demonstrates that the travel time estimates are very good provided there are no sources of delay, such as an incident, within a link.

Day-to-Day Travel-Time Trends and Travel-Time Prediction from Loop-Detector Data

An approach is presented for estimating future travel times on a freeway using flow and occupancy data from single-loop detectors and historical travel-time information. Linear regression, with the stepwise-variable-selection method and more advanced tree-based methods, is used. The analysis considers forecasts ranging from a few minutes into the future up to an hour ahead. Leave-a-day-out cross-validation was used to evaluate the prediction errors without underestimation. The current traffic state proved to be a good predictor for the near future, up to 20 min, whereas historical data are more informative for longer-range predictions. Tree-based methods and linear regression both performed satisfactorily, showing slightly different qualitative behaviors for each condition examined in this analysis. Unlike preceding works that rely on simulation, real traffic data were used. Although the current implementation uses measured travel times from probe vehicles, the ultimate goal is an autonomous system that relies strictly on detector data. In the course of presenting the prediction system, the manner in which travel times change from day to day was examined, and several metrics to quantify these changes were developed. The metrics can be used as input for travel-time prediction, but they also should be beneficial for other applications, such as calibrating traffic models and planning models.

IMPROVED VELOCITY ESTIMATION USING SINGLE LOOP DETECTORS

This paper develops an improved algorithm for estimating velocity from single loop detector data. Unlike preceding works, the algorithm is simple enough that it can be implemented using existing controller hardware. The discussion shows how the benefits of this work extend to automated tests of detector data quality at dual loop speed traps. Finally, this paper refutes an earlier study that found conventional single loop velocity estimates are biased.

The PeMS algorithms for accurate, real-time estimates of g-factors and speeds from single-loop detectors

Presents the PeMS algorithms for the accurate, adaptive, real-time estimation of the g-factor and vehicle speeds from single-loop detector data. The estimates are validated by comparison with independent, direct measurements of the g-factor and vehicle speeds from 20 double-loop detectors on I-80 over a three-month period. The algorithm is used to process data from all freeways in Caltrans District 12 (Orange County, CA) over a 20-month period beginning January 1998. Analysis of those data shows that the g-factors for different loops in the district differ by as much as 100 percent, and the g-factor for the same loop can vary up to 50 percent over a 24-hour period. Many transportation districts now post real-time speed and travel time estimates on the World Wide Web. Those estimates often are derived from single-loop detector data assuming a common g-factor for all detectors in the district. This study suggests that those estimates can be in error by 50 percent, and so they are of little value to travelers. The use of the PeMS algorithm will reduce those errors.

Vehicle reidentification and travel time measurement on congested freeways

Abstract The paper presents an algorithm for matching individual vehicles measured at a freeway detector with the vehicles’ corresponding measurements taken earlier at another detector located upstream. Although this algorithm is potentially compatible with many vehicle detector technologies, the paper illustrates the method using existing dual-loop detectors to measure vehicle lengths. This detector technology has seen widespread deployment for velocity measurement. Since the detectors were not developed to measure vehicle length, these measurements can include significant errors. To overcome this problem, the algorithm exploits drivers’ tendencies to retain their positions within dense platoons. The otherwise complicated task of vehicle reidentification is carried out by matching these platoons rather than individual vehicles. Of course once a vehicle has been matched across neighboring detector stations, the difference in its arrival time at each station defines the vehicle’s travel time on the intervening segment. Findings from an application of the algorithm over a 1/3 mile long segment are presented herein and they indicate that a sufficient number of vehicles can be matched for the purpose of traffic surveillance. As such, the algorithm extracts travel time data without requiring the deployment of new detector technologies. In addition to the immediate impacts on traffic monitoring, the work provides a means to quantify the potential benefits of emerging detector technologies that promise to extract more detailed information from individual vehicles.

USING DUAL LOOP SPEED TRAPS TO IDENTIFY DETECTOR ERRORS

Dual loop speed traps have a distinct advantage over single loop detectors because the speed trap detection system is redundant. Each vehicle is observed twice under normal operating conditions, once at each loop. The two observations are normally used to measure velocity, but as this paper demonstrates, the redundancy can also be used to assess the performance of the speed trap and identify detector errors. At free-flow velocities, the time each detector is occupied by a vehicle (i.e., the ontime) should be virtually identical, regardless of the vehicle length. Many hardware errors will cause the two on-times to differ. Exploiting this property, a formal methodology for testing speed traps off-line has been developed, and ways to extend the work to on-line testing are suggested. The work is used to evaluate several loop sensor units, revealing problems in two models. A second example shows how the work can be used to detect cross talk between sensor units.

ESTIMATING MEDIAN VELOCITY INSTEAD OF MEAN VELOCITY AT SINGLE LOOP DETECTORS

Loop detectors are the preeminent vehicle detector for freeway traffic surveillance. Although single loops have been used for decades, debate continues on how to interpret the measurements. Many researchers have sought better estimates of velocity from single loops. The preceding work has emphasized techniques that use many samples of aggregate flow and occupancy to reduce the estimation error. Although rarely noted, these techniques effectively seek to reduce the bias due to long vehicles in measured occupancy. This paper presents a different approach, using a new aggregation methodology to estimate velocity and reduce the impact of long vehicles in the original traffic measurements. In contrast to conventional practice, the new estimate significantly reduces velocity estimation errors when it is not possible to control for a wide range of vehicle lengths.

Vehicle Re-Identification and Travel Time Measurement in Real-Time on Freeways Using Existing Loop Detector Infrastructure

A new vehicle re-identification algorithm for two consecutive detector stations on a freeway, whereby a vehicle measurement made at the downstream detector station is matched with the vehicle’s corresponding measurement at the upstream station, is presented in this paper. The method is illustrated using effective vehicle length measured at dual-loop speed traps, but it is transferable to other detectors capable of extracting a vehicle signature (such as video image processing). This approach is significant because no one has attempted to use the existing detector infrastructure to match vehicle measurements between detector stations. The algorithm should improve freeway surveillance via travel time measurement, which is simply the difference between the known arrival times at the two stations for a matched vehicle. The re-identification algorithm is tolerant to noise; instead of finding the ‘best match’ for each vehicle, it finds all possible matches and then looks for sequences of vehicles from the possible matches. Even with noisy loop detector data, the sequence detection eliminates most of the possible-but-incorrect matches while the true matches remain. The new methodology will be used to examine the applications and benefits of travel-time data on real-world traffic, without the expensive costs of installing new detectors. Ordinarily, a travel-time measurement system would have to be fully deployed before the benefits can be quantified.