Welcome to Prof. Taek Kwon's Research Page

This page is intended to provide a brief overview of my current research work at UMD. Unfortunately, it seems, I could never find enough time to construct a good web page that I wish to provide. It will be always partial, but if anyone is interested in my research beyond what is available here, please don't hesitate to contact me. 218-726-8211.

Selected Research Reports

Rare Papers frequently requested

My Recent Research Projects:

Develepment And Evaluation Of An Advanced LED Warning System For Rural Intersections

In Minnesota, 70 percent of all intersection-related fatal crashes for the period of 2006 to 2008 occurred at rural, through/stop intersections. The Minnesota Department of Transportation (Mn/DOT) identified improving the design and operation of intersections as a critical emphasis area in the Minnesota Comprehensive Highway Safety Plan. At these intersections, sight restrictions caused by vertical and horizontal curves negatively affect a driver’s ability to safely accept a gap in the traffic stream. This paper presents the result of a two year study on the development of a new intersection warning system referred to as an Advanced Light Emitting Diode (LED) Warning System (ALWS) and evaluation of the system’s effectiveness. The ALWS was developed to address the sight restrictions in rural through/stop intersections, and it consists of vehicle detectors that detect approaching or stopped vehicles and LED warning signs that respond according to the received messages from the detectors. The warning signs have LEDs on the perimeter of the sign and a warning message in the middle, which is commonly referred to as a blinker sign. All signs and detectors are powered by solar panels and rechargeable batteries. All message communications between the detectors and warning signs are performed through wireless transceivers. In order to evaluate this new warning sign technology, the system was installed at the intersection of West Tischer Road and Eagle Lake Road in Duluth, Minnesota. Video data was collected through an on-site video monitoring system consisting of a Digital Video Recorder (DVR) and two video cameras. The first camera records video of vehicles traveling towards the intersection through the vertical curve. The second camera records vehicles traveling through the intersection. In addition to video data collection, mail-in and on-site surveys were conducted. Overall, the ALWS was effective at reducing vehicle speeds on the main approach, and increasing the wait time and altogether stopping roll-throughs for vehicles on the minor approaches when a conflict exists at the intersection. However, an increase in roll-throughs when no conflict exists at the intersection was observed, which must be addressed in the future design of the ALWS. According to the mail-in and on-site survey results, 80 percent of respondents expressed that the warning system is effective. If the increase in roll-throughs under no conflict conditions can be addressed, the researchers conclude that the ALWS is an effective system for reducing crashes in rural stop/through intersections.
Youtube Video Demo: a Working Sign.

Development of a Weigh-Pad Based Portable WIM System

Since mechanistic designs were introduced in the 2002 Design Guide, weigh-in-motion (WIM) data is required as one of the primary inputs to pavement design. Also, recent increases in heavy truck volumes on local roads caused by higher demands on agricultural commodities raise great concerns on the life of existing local road infrastructure, elevating the needs of truck weighing and enforcements. Installing many WIM stations could address these needs. However, the cost for installing a permanent WIM station is expensive (over $200,000) and requires recurring costs of maintenance trips, electricity, and communication. For county roads with limited average daily traffic (ADT), such a high cost installation and maintenance is rarely justifiable. One solution to brining the WIM technologies to local roads is to use low-cost portable WIM systems that can be easily moved around. Unfortunately, portable and easily deployable WIM systems are presently not available on the market. This research project develops a new weigh-pad based portable WIM system that can be easily installed and used much like a tube counter. The developed system will be battery operated, low cost, and easily installable on both rigid and flexible pavements. An additional benefit of the weigh pad based WIM system is that, since it does not cut into the pavements during the installation, it can be safely used on structurally sensitive areas such as on bridge decks.

Cellular Wireless Mesh Sensor Network for Vehicle Tracking in an Intersection

Automatic vehicle tracking has been one of the challenges in traffic detection technologies. Video cameras and radars had only limited success in part due to the problems with occlusion, land object noises, weather, and high computational and physical costs. This research project involves the development of a reliable cellular wireless mesh sensor network (WMSN), utilizing recent advances in ZigBee technology. The WMSN is self-constructing, self-healing, and can support a large number of nodes. Each node of the WMSN has a minimum footprint that consists of a microcontroller with a radio frequency transceiver, an anisotropic magnetoresistance sensor for detecting vehicles, and a lithium-ion rechargeable battery. A node is placed in each lane of the intersection to form a WMSN. A separate node is responsible for collecting and logging the data from each node in the network. From this logged data, a vehicle tracking algorithm analyzes the logged intersection data and tracks the trajectories of the vehicles through the intersection. Actual intersection data was collected and successfully tracked using the vehicle tracking algorithm with an average of over 90 percent accuracy. Final Report: Portable Cellular Wireless Mesh Sensor Network for Vehicle Tracking in an Intersection

Development of a PC-Based Eight-Channel WIM System

Weigh-in-Motion (WIM) data provides vital information for pavement design and maintenance. The purpose of this research project was to improve the present piezoelectric WIM technologies through a better system design and signal processing algorithms. Present WIM systems are only available as proprietary systems, i.e., the internal system design and algorithms are highly guarded making it difficult to compare and improve the underlying technology. Therefore, the second objective was to develop a WIM system based on an open architecture, utilizing a standard PC and off-the-shelf components, and to publish the details of the design to promote an open architecture for continuous future improvements by other developers. The research team was able to successfully develop a working eight-channel WIM system, and the details are described in this report. The main innovation introduced in this research is a hardware-in-the-loop (HIL) WIM simulator that can generate analog axle and loop signals through software control. The HIL simulator can create ideal axle signals, as well as erroneous signal conditions, that can be directly fed into WIM systems. The main advantage of using a WIM HIL simulator for developing a WIM system is that the developers may run an unlimited number of signal tests without actually driving a single vehicle through the WIM sensors, thereby significantly reducing the development time and cost. The erroneous signal conditions generated by the HIL simulator can also identify the error handling capabilities of a WIM system. The proposed HIL simulator for WIM system development is new and provides an elegant solution to the unavailability of an ideal axle signal.

Visibility Measurement System Based on Imaging Technologies

This research project focuses on developing a practical atmospheric-visibility monitoring system based on imaging systems. Because visibility reductions due to inclement weather conditions are one of the main causes of traffic incidents, and among one of the primary criteria used to determine road closures in winter, accurate visibility measurement is prime importance to many transportation decision makers. However, accurate and reliable measurement of atmospheric visibility is challenging because it continuously changes over time and space and is influenced by a host of atmospheric conditions such as fog, rain, snow, smog, dust, sun direction, solar radiation, etc. Measurement by human observers is often unreliable due to differences in individual eyesight, perception and other biological conditions. Other techniques such as light-scatter meters exist, but do not measure the true visibility. A new approach developed in this research is based on the measurement of visibility through an imaging system that comprises a surveillance video camera, an image digitizer and multiple targets positioned at specific distances from the camera. For daytime, an image-processing algorithm was developed to determine at what distance the foreground is no longer distinguishable, by which the distance represents the visibility. For night visibility, several approaches are under investigation. Those include embedding light sources in the targets with the frequency range in visual and infrared spectrum and using different types of spectral filters in the camera. We also investigate a new way of measuring visibility based on the concept of relativity. This research was sponsored by US DOT TEA-21 and Mn/DOT. $93,191.
Click here to see a daytime visibility screen of the system.
Click here to see a night visibility screen of the system.
Final report Phase-1: Atmospheric Visibility Measurements Using Video Cameras
Final report Phase-2: Atmospheric Visibility Measurements Using Video Cameras: Relative Visibility

The following areas of research are not funded, but under way due to my own interests.

Global Optimization Neural Network

One-Pass Learning Neural Network

Non-Uniform Image Processing

Most of the techniques employed in the area of image processing have been explored by uniform processing approaches in its nature. For example, image compression uniformly compresses all areas of the image without regarding such as the human perception of objects in the image (which may provide critical information on "which portion of the image is more interested or important than others"). Another such area is in image interpolation. Most interpolation techniques apply one formulation or another without knowing how much distortion would be caused to the image as a result of this uniform formulation of the interpolation algorithm, which essentially leads to introducing many blocky effects. The effect on the Fourier Transform is a non-issue and at times it is very cumbersome to find the effects of these arbitrary interpolation techniques on the frequency domain. This creats havoc on the frequency components. A better and sensible approach is associating a non-uniform processing based on human perception. My research effort is to develop a non-uniform image processing that accounts human perception in particular for image compression and interpolation.

Object Recognition

The area of object recognition has traditionally been synonymous with the technique of template matching. However I believe that it is highly unreliable due to some of its inherent characteristics. My effort is to develop a newer technique that can be more generalized in terms of human intelligence and perception. The direction of this research is to follow the simple common-sense concept that we human beings detect or recognize a pattern or an object from a scene by interpreting the syntax of features that make up an object. This approach is incrediably efficient in terms of data as demonstrated by few strokes of cartoon artists. Unfortunately, we know very little about how to automatically extract features and feature-syntax relation from raw data. I am presently in the process of developing few new techiques and applying the algorithm in visibility measurements for the Deaprtment of Transportation.

Back to my home page Kwon's home Page.

Back to the page2 page