Doctoral Dissertations

Date of Award

8-2007

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Industrial Engineering

Major Professor

John C. Hungerford

Committee Members

Tyler Kress, Kenneth E. Kirby, Lee D. Han, Stephen H. Richards

Abstract

The aim of this research is to utilize PhotoModeler, a closerange photogrammetry software package, in various traffic accident reconstruction applications. More specifically, three distinct studies were conducted: 1.) vehicle crush measurement, 2.) road curve measurement, and 3.) an evaluation of common traffic accident reconstruction measurement methodologies.

The first study applied the photogrammetric process to controlled crash information generated by the National Highway Traffic Safety Administration (NHTSA). A statistical procedure known as bootstrapping was utilized to generate distributions from which the variability was examined. The “within” subject analysis showed that 44.8% of the variability is due to the technique itself and the “between” subjects analysis demonstrated that 55.2% of the variability is attributable to vehicle type—roughly half and half. Additionally, a 95% CI for the “within” analysis revealed that the mean difference (between this study and NHTSA) fell between -2.52 mph and +2.74 mph; the “between” analysis showed a mean difference between -3.26 and +2.41 mph.

The second study focused on photogrammetry in road curve measurement. More particularly, this study applied photogrammetry to (simulated) road curves in lieu of traditional measurement methods, such as measuring tapes and measuring wheels. In this work, thirty (30) different radii of curvature of various known sizes were deliberately constructed. Then photogrammetry was used to measure each of the constructed curves. A comparison of the known “R’s” (control group) and photogrammetry’s value of “R” (treatment group) was then made. Matched Pairs or Paired Comparisons were then used to examine these two populations. The difference between the photogrammetry “R” and the known “R” range is between 0.001% and 0.874%. Additionally, we are 95% confident that the mean difference of the two techniques is between -0.33 and 0.51 feet. Since this interval contains zero, we can conclude that the two techniques do not differ.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS