Determination of vehicle origins and destinations have long been a goal and a tool for traffic planners. Recently there has been an increase in the use of machine vision systems (automatic license plate readers) for determining origins and destinations. The original work in this area involved recording traffic on s-video compatible tape formats (e.g. s-VHS and Hi-8mm). Recently the use of real time analysis of traffic flows has been implemented. This paper describes the implementation of a system to acquire license plate information on the Tacoma Narrows Bridge, in Washington state. The paper describes the system installation, the system architecture, the data handling and mail outs and some preliminary results.

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 and experiments of our tracker/grouper on several minutes of videotape.

Automobile machine vision is a potential application for hardware neural networks. This paper suggests a future system concept which combines front-end low-level processing in smart pixels and a massively interconnected optical neural network. Two smart pixels devices fabricated in Ferroelectric Liquid Crystal on Silicon technology will be described. The first device implements an array of somas, while the second device implements an array of synapses. These are effectively generic building blocks for scalable neural network architectures. Experiments have demonstrated actions of optoelectronic neurons formed by the smart pixels. An optical neural network system has been investigated in a simulation of an autoassociative memory for road sign recognition.

A major obstacle in the application of stereo vision to intelligent transportation system is high computational cost. In this paper, a PC based three-camera stereo vision system constructed with off-the-shelf components is described. The system serves as a tool for developing and testing robust algorithms which approach real-time performance. We present an edge based, subpixel stereo algorithm which is adapted to permit accurate distance measurements to objects in the field of view using a compact camera assembly. Once computed, the 3D scene information may be directly applied to a number of in-vehicle applications, such as adaptive cruise control, obstacle detection, and lane tracking. Moreover, since the largest computational costs is incurred in generating the 3D scene information, multiple applications that leverage this information can be implemented in a single system with minimal cost. On-road applications, such as vehicle counting and incident detection, are also possible. Preliminary in-vehicle road trial results are presented.

In the 1995 Collision Avoidance and Automated Traffic Management Sensors Proceedings we presented a stereo vision based approach for detecting the relative location of other vehicles in highway driving. This system was able to track other vehicles which were within the stereo pair visual field of view. In addition, important camera position variables such as inclination angle and camera height were dynamically calculated from image data without the need for external sensors. In collaboration with industrial partners we have tested the algorithm on additional video information under a number of different driving situations. The affect of noise in the camera geometry estimate on the algorithm's performance was also examined. As a result of these tests and the additional requirement of real time performance, a number of modifications were made to the algorithm to increase both its reliability and computational efficiency. We present the results of the testing as well as the algorithmic changes here.

MIT Lincoln Laboratory is developing new electronic night vision technologies for defense applications which can be adapted for civilian applications such as night driving aids. These technologies include (1) low-light CCD imagers capable of operating under starlight illumination conditions at video rates, (2) realtime processing of wide dynamic range imagery (visible and IR) to enhance contrast and adaptively compress dynamic range, and (3) realtime fusion of low-light visible and thermal IR imagery to provide color display of the night scene to the operator in order to enhance situational awareness. This paper compares imagery collected during night driving including: low-light CCD visible imagery, intensified-CCD visible imagery, uncooled long-wave IR imagery, cryogenically cooled mid-wave IR imagery, and visible/IR dual-band imagery fused for gray and color display.

A method is developed and investigations of the contrast and limiting visibility range of a retroreflective marker defining, for example, the overall-dimensions of a heavy truck are carried out for a case of illumination by anti-fog headlamps, upper or low beam of another vehicle through fog of various optical density with no other external lighting. The studies simulate night vision of non-Lambertian road markers from the viewpoint of the modern image transfer theory. The method enables one to include analytically the effects of multiple light scattering by fog aerosols, different mutual positions of the headlamps, vehicle driver, and object observed, as well as real retroreflective properties of the marker to derive vision characteristics via quite simple formulas. The visual contrast and limiting visibility range values of, respectively, retrorefractive and Lambertian markers are compared to demonstrate the advantages of applying the former for enhancing traffic safety and increasing allowable road speed. The recommendations are made with respect to the use of retroreflective markers as an auxiliary signaling means to show the overall dimensions.

A laboratory prototype of a passive optical lane position monitor has been designed, built, and tested. The sensor head is simple and consists of two parts: a cylindrical lens, and a position sensitive detector. The amplifier/processing electronics which provides the position signal is compact and lightweight. No complex software or computer is needed. Sensor performance was validated both in the laboratory and in the field. The prototype was tested in sunlight over a range of solar angles from dawn to dusk. It was even tested at night with illumination provided by headlights. The bottom line is that, for such a simple system, the sensor worked quite well. This opens up possibilities for its use as a practical tool in vehicle/highway management.

Aerometrics initiated and continues on the development an innovative laser-diode based device that provides a warning signal when a motor-vehicle deviates from the center of the lane. The device is based on a sensor that scans the roadway on either side of the vehicle and determines the lateral position relative to the existing painted lines marking the lane. The principles of operation of the sensor, and the results of Aerometrics' early testing were presented last year in this forum. This paper presents Aerometrics' continuing efforts in bringing the technology to market. New prototypes have been developed and tested. Aerometrics' engineering efforts and the use of latest technologies have resulted in a 24-fold reduction in sensor volume when compared to their predecessors and similar reductions in weight. The current prototype measures less than 9 cm X 8 cm X 7 cm, and can be easily fit within the cavity of rear-view mirror holders used in most present-day vehicles. Also, advances in signal conditioning and processing have improved the reliability of the sensor. Results of continuing testing of the sensor will be presented.

Single vehicle run-off-road accidents are responsible for significant numbers of injuries and fatalities, and significant property damage. This fact spurs interest in warning systems to alert drivers that vehicles are drifting towards the edge of the road, and that a run-off road accident is imminent. An early attempt at such a warning system is the use of machined grooves on the shoulder to create a rumble strip. Such a system only provides warning, however, as the vehicle actually leaves the traffic lane. More desirable is a system that warns in anticipation of such departure. Honeywell has under development a magnetic lateral guidance system that couples a sensitive magnetoresistive transducer with a magnetic traffic marking tape being developed by 3M. While this development was initially undertaken for use in automated highways, or for special tasks such as guiding snowplow owners, the system can provide an effective, all-weather warning system to provide alert of impending departure from the roadway. This electronic rumble strip is actually a simpler system than the baseline guidance system, and can monitor both distance from the traffic lane edge and the speed of approach to the edge with a low cost sensor.

This paper discusses a configuration of overhead active- infrared imaging vehicle sensors that can be used to monitor traffic on freeways and their entrance and exit ramps in order to provide the information needed to optimize the flow of traffic. The Wide-Area Traffic-Surveillance system, which is being developed under a program with the Jet Propulsion Laboratory, comprises single-lane and three-lane sensors that employ line-scanned laser rangefinders to measure the presence, speed, and 3D profiles of vehicles passing beneath them in single-lane or three-lane fields of view, respectively. The time-tagged outputs of the various sensors are routed, via hard wire or radio link, to a central processor where site-specific software is used to generate a computer image of the real-time area-wide traffic flow, with icons representing vehicles. The software is also used to determine traffic parameters including vehicle flow rate, average speed, density, classification, and travel time as well as ramp demand, passage, and queue length. Principles of sensor operation, system architecture, hardware and software functions, and interfaces and communication protocols will be described. Test results obtained at two sites on a major arterial in Orlando, Florida, will be presented.

A vision-based traffic surveillance processor is being developed at JPL. This processor uses innovative image segmentation and classification techniques for vehicles in freeway images, including those with large shadows. These results enable the computation of many useful traffic parameters. A wavelet-based algorithm has been developed for vehicle detection and segmentation. Specifically, two types of mother wavelet has been created and tested: the first for shape-size discrimination of vehicles from their background; and the second for locating where vehicles join their shadows, thus enabling segmentation of the vehicles from their shadows. Combining these two wavelets enables robust segmentation of vehicles from busy freeways. This method reduces the false-alarm rate in vehicle counts, since shadows are no longer mistaken for vehicles. We use neural networks for vehicle classification. To reduce system complexity and training time, we use, as preprocessors, several feature extraction methods, such as invariant-moment and Hermite-moment computations. This preprocessing enables orders of magnitude reductions in training time and a great increase in classification accuracy.

This paper evaluates the accuracy of three color categorization techniques in describing vehicles colors for a system, AutoColor, which we are developing for Intelligent Transportation Systems. Color categorization is used to efficiently represent 24-bit color images with up to 8 bits of color information. Our inspiration for color categorization is based on the fact that humans typically use only a few color names to describe the numerous colors they perceive. Our Crayon color categorization technique uses a naming scheme for digitized colors which is roughly based on human names for colors. The fastest and most straight forward method for compacting a 24-bit representation into an 8-bit representation is to use the most significant bits (MSB) to represent the colors. In addition, we have developed an Adaptive color categorization technique which can derive a set of color categories for the current imaging conditions. In this paper, we detail the three color categorization techniques, Crayon, MSB, and Adaptive, and we evaluate their performance on representing vehicle colors in our AutoColor system.

In order to support intelligent traffic management systems, future traffic sensors will need to provide more traffic variables, and more accurate data than provided by the commonly used sensors today. In particular, future systems will need to recognize vehicles from sensor to sensor in order to measure point-to-point traffic variables like travel time. One characteristic that may be used to match vehicles between sensors is color. This work describes a machine vision based traffic sensor under development that uses sophisticated color signatures as a key component.

An important part of traffic management and intelligent transportation systems is the availability of cost effective, reliable sensors for vehicle classification and traffic flow monitoring. Fiber optic sensors have numerous advantages over conventional sensors and are thus prime candidates for use in these traffic systems. In this paper, an optical fiber vehicle sensor is described for axle detection. Results from field tests are presented. The performance of the fiber optic vehicle sensor is evaluated and methods of vehicle weight-in-motion are also discussed.

We describe new fiber-optic vibrational and acoustic sensor systems developed for airport ground traffic monitoring. The theoretical background is derived and results of laboratory experiments as well as initial field tests in an Experimental Surface Movement Guidance and Control System at the Braunschweig airport are reported.

New technologies are being used to provide solutions to old problems in many areas of our society, including the transportation industry. Congestion on highway systems around the world is a growing concern because of the great expense it causes to the transportation industry through time delays and increased operating costs. The cost of congestion has been estimated to be worth billions of dollars annually and causes a great drain on the economy. The development of monitoring technology will provide the ability to address the congestion problem in new and economical ways. This paper discusses the development and applications of a non-intrusive detection technology using passive acoustic sensors. Applications of this non-intrusive technology include traffic management and surveillance for major metropolitan areas, and traffic data collection based on vehicle count, speed, and classification. The SmartSonic^TM passive acoustic detector provides the basic information necessary for a Freeway management system. A summary will be presented of the basic technology and operation of the SmartSonic^TM system, and will examine the results of independent testing performed to evaluate the performance of the system. Independent evaluation indicates that the SmartSonic^TM system is a good alternative to traditional inductive loop technology in applications such as freeway management, providing easy installation and reliable information.

The aim of this research is to investigate the feasibility of developing a cost effective traffic monitoring detector for the purpose of reliable on-line vehicle classification to aid traffic management systems. The detector used was a directional microphone connected to a DAT recorder. The digital signal was preprocessed by LPC (Linear Predictive Coding) parameter conversion based on autocorrelation analysis. A Time Delay Neural Network (TDNN) was chosen to classify individual travelling vehicles based on their speed-independent acoustic signature. The network was trained and tested with real data for four types of vehicles. The paper provides a description of the TDNN architecture and training algorithm and an overview of the LPC pre-processing and feature extraction technique as applied to audio monitoring of road traffic. The performance of TDNN vehicle classification, convergence and accuracy for the training patterns are fully illustrated. In generalizing to a limited number of test patterns available, 100% accuracy in classification was achieved. The net was also robust to changes in the starting position of the acoustic waveforms with 86% accuracy for the same test data set.

This paper describes two fiber Bragg grating based strain sensing systems. The two systems are designed for separate applications, one for many-sensor static strain measurements while the other is capable of monitoring strains with high sampling rates. Results are presented for the use of the dynamic system for monitoring strain on an active I-10 bridge in New Mexico.

Accurate and reliable detection, discrimination, and quantitative measurement of visibility conditions will be crucial in the implementation of weather advisory schemes for Intelligent Transportation Systems. Increasing the amount of information about inclement weather to the motorist can greatly increase the levels of safety and mobility, reduce congestion, and enhance overall economic productivity. Typically, the number of visibility sensors that are deployed along the highway is limited due to their high-cost. Poor visibility conditions, however, are not uniform along the length of the highway, and it is therefore important that a larger number of low-cost and reliable sensors are deployed. Optical fiber sensors which possess unique advantages as compared to conventional electrical sensors may provide an interesting alternative to this problem.

The advanced remote vehicle emission sensing equipment, Smog Dog, is a cost-effective infrared technology designed to measure the levels of vehicle exhaust. This paper presents and demonstrates a research effort for using the Smog Dog to conduct the on-road vehicle exhaust emission collection in the city of Houston, develop modal sensitive emission models and evaluate the EPA approved MOBILE5A emission factor model. The vehicle emission data collection is designed in a manner that various vehicle's modal events such as the acceleration and deceleration under the on-road driving conditions are considered. The Smog Dog remote mission sensor can not only collect the emission concentrations of hydrocarbon, carbon monoxide and oxide of nitrogen but also simultaneously detect the vehicles' instantaneous speeds and acceleration rates. Thus a vehicle's emission rates, which are converted from the collected emission concentration levels, can be functions of its instantaneous speed and acceleration rate. In addition, the Federal Test Procedure driving cycles are emulated using the emission versus speed profile relationships and the resulted emission rate for a predetermined average driving speed can then be compared with the emission factors produced by MOBILE5A. Since the emission models, that are developed based on the on-road emission data collected using the Smog Dog, naturally reflect the on-road driving conditions and the vehicle fleet combinations, they can potentially be used to evaluate the vehicle exhaust emission implications of various advanced traffic management strategies.

Under a program sponsored by the Department of Energy, the Oak Ridge complex is developing a `Portal-of-the-Future', or `smart portal.' This is a security portal for vehicular traffic which is intended to quickly detect explosives, hidden passengers, etc. It uses several technologies, including microwaves, weigh-in-motion, digital image processing, and electroacoustic wavelet-based heartbeat detection. A novel component of particular interest is the Enclosed Space Detection System (ESDS), which detects the presence of persons hiding in a vehicle. The system operates by detecting the presence of a human ballistocardiographic signature. Each time the heart beats, it generates a small but measurable shock wave that propagates through the body. The wave, whose graph is called a ballistocardiogram, is the mechanical analog of the electrocardiograms, which is routinely used for medical diagnosis. The wave is, in turn, coupled to any surface or object with which the body is in contact. If the body is located in an enclosed space, this will result in a measurable deflection of the surface of the enclosure. Independent testing has shown ESDS to be highly reliable. The technologies used in the smart portal operate in real time and allow vehicles to be checked through the portal in much less time than would be required for human inspection. Although not originally developed for commercial transportation, the smart portal has the potential to solve several transportation problems. It could relieve congestion at international highway border crossings by reducing the time required to inspect each vehicle while increasing the level of security. It can reduce highway congestion at the entrance of secure facilities such as prisons. Also, it could provide security at intermodal transfer points, such as airport parking lots and car ferry terminals.

Insects tend to detect motion rather than images and this together with inherent parallelism in their visual architecture, leads to an efficient and compact means of collision avoidance. A VLSI implementation of a smart microsensor that mimics the early visual processing stage in insects has been developed. The system employs the `smart sensor' paradigm in that the detectors and processing circuitry are integrated on one chip. The IC is ideal for motion detectors, particularly collision avoidance tasks, as it essentially detects the speed, bearing and time-to-impact of a moving object. The Horridge model for insect vision has been directly mapped into VLSI and therefore the IC truly exploits the beauty of nature in that the insect eye is so compact with parallel processing, enabling compact motion detection without the computational overhead of intensive imaging, full image extraction and interpretation. This world-first has exciting applications in areas such as anti- collision for automobiles and autonomous robots.

A research program oriented toward the development of a portable data acquisition system for crash avoidance research has been conducted. This paper discusses the background to the project and the requirements for the data acquisition system. It also provides a brief system overview and describes two of the system's five major elements, the sensor suite and the video data system, in detail. Components, functions, and specifications are covered. Finally the paper addresses the central data collection/analysis facility which was assembled to mange the sensor and video data, and presents the potential uses of the data acquisition system.

On-vehicle sensors for collision avoidance and intelligent cruise control are receiving considerably attention as part of Intelligent Transportation Systems. Most of these sensors are radars and `look' in the direction of the vehicle's headway, that is, in the direction ahead of the vehicle. Calspan SRL Corporation is investigating the use of on- vehicle radar for Intersection Collision Avoidance (ICA). Four crash scenarios are considered and the goal is to design, develop and install a collision warning system in a test vehicle, and conduct both test track and in-traffic experiments. Current efforts include simulations to examine ICA geometry-dependent design parameters and the design of an on-vehicle radar and tracker for threat detection. This paper discusses some of the simulation and radar design efforts. In addition, an available headway radar was modified to scan the wide angles (+/- 90 degree(s)) associated with ICA scenarios. Preliminary proof-of-principal tests are underway as a risk reduction effort. Some initial target detection results are presented.

A generalized methodology for evaluating alerting systems is presented. The methodology is used to construct System Operating Characteristic curves that describe the tradeoffs between unnecessary alerts and collisions based on probabilistic models of sensors, vehicle dynamics, and human response. An appropriate alerting threshold location can then be selected and parametric studies can be performed to examine the relative benefits of alternative sensor types and accuracies. Different avoidance maneuver options (e.g., swerving or braking) can be evaluated in terms of their ability to reduce the probability of a collision. A simplified example application is presented for a ground vehicle rear-end collision alerting system. The relative benefits of increased sensor accuracy vs. improved driver response time and braking deceleration are examined. It is shown that uncertainty in human response time is the key factor affecting the performance of the alerting system. In contrast, sensor accuracy specifications are shown to be entirely adequate: system performance is not significantly impacted by expected sensor errors.

Intersection collisions constitute approximately twenty-six percent of all accidents in the United States. Because of their complexity, and demands on the perceptual and decision making abilities of the driver, intersections present an increased risk of collisions between automobiles. This situation provides an opportunity to apply advanced sensor and processing capabilities to prevent these collisions. A program to determine the characteristics of intersection collisions and identify potential countermeasures will be described. This program, sponsored by the National Highway Traffic Safety Administration, utilized accident data to develop a taxonomy of intersection crashes. This taxonomy was used to develop a concept for an intersection collision avoidance countermeasure. The concept utilizes in-vehicle position, dynamic status, and millimeter wave radar system and an in-vehicle computer system to provide inputs to an intersection collision avoidance algorithm. Detection of potential violation of traffic control device, or proceeding into the intersection with inadequate gap will lead to the presentation of a warning to the driver. These warnings are presented to the driver primarily via a head-up display and haptic feedback. Roadside to vehicle communication provides information regarding phased traffic signal information. Active control of the vehicle's brake and steering systems are described. Progress in the development of the systems will be presented along with the schedule of future activities.

There are approximately 20 million police radar detectors used on the highways of the United States daily. A highway hazard safety warning system has been developed by the Georgia Tech Research Institute, working under the sponsorship of the radar detector industry, to communicate highway safety alerts to the driver of any vehicle equipped with a police radar detector. In addition, the system causes the new generation of detectors that are already available to display a safety warning message on an alpha-numeric display. The Safety Warning System^TM consists of a transmitter and a radar detector receiver or stand-alone safety warning receiver/display system. The transmitter can be mounted on police cars, emergency vehicles, utility vehicles, highly repair vehicles, and on stationary structures at fixed locations along the highway. The reception range of the transmitted signal is between 0.5 and 1.0 miles, depending on terrain. The system to be described may be one of the first applications of in-vehicle signing in the Intelligent Transportation System to be implemented, because the required infrastructure of receivers already exists.

Three experiments were performed with examined the applicability of the Hick-Hyman law to the design of an auditory interface for a vehicle collision avoidance warning system. All trials used a single broadband noise signal emanating from one of a subset of six loudspeakers equally spaced around the subject in the azimuthal plane. Both the size of the sub-set and the balance of relative probabilities from speaker to speaker were altered to evaluate the relationship between information content and the dependent variable, choice reaction time. Choice reaction time was found to be related to the information content of the sound stimulus in all cases. It was also found to be related to the presence of pairs of speakers which were symmetrically opposed to one another in front of and behind the subject.

Video systems can locate, identify and track vehicles. A video-based vehicle detection and location method is presented, which exploits the symmetry of vehicles seen from behind. The method can account for a range of symmetry types. These include (1) simple pixel presence in a binary edge image, (2) gray level of the edge pixel, (3) color value of the edge pixel, and (4) connectedness structure of the (binary) pixels around an edge pixel. A fast algorithm for the generation of a symmetry histogram is presented, whose (sufficiently strong) peaks indicate the likely presence and approximate location of a vehicle. The speed of the algorithm results from its data driven nature. Typical results for this algorithm in dense urban traffic are presented, using symmetry type (1). The generalization of the algorithm to skew symmetric images is shown. A potential application of the algorithm to automated roadways and fatigue detection is sketched out. Robustifying extensions in the spirit of (2), (3) and (4) are proposed.

On-vehicle technology for collision avoidance using millimeter wave radar is currently under development and is expected to be in vehicles in coming years. Recently approved radar bands for collision avoidance applications include 47.5 - 47.8 GHz and 76 - 77 GHz. Widespread use of active radiation sources in the public domain would contribute to raised levels of human exposure to high frequency electromagnetic radiation, with potential for adverse health effects. In order to design collision avoidance systems that will pose an acceptably low radiation hazard, it is necessary to determine what levels of electromagnetic radiation at millimeter wave frequencies will be acceptable in the environment. This paper will summarize recent research on NIR (non-ionizing radiation) exposure safety standards for high frequency electromagnetic radiation. We have investigated both governmental and non- governmental professional organizations worldwide.