Type of Document	Thesis
Author	Mauga, timur Ibrahim
Author's Email Address	timurmauga@yahoo.com
URN	etd-05052006-115118
Title	The Development of Florida Length Based Vehicle Classification Scheme Using Support Vector Machines
Degree	Master of Science
Department	Civil and Environmental Engineering, Department of
Advisory Committee	Advisor Name Title Renatus Mussa Committee Chair John Sobanjo Committee Member Wei-Chou Virgil Ping Committee Member
Keywords	Support Vector Macines Pattern Recognition Scheme F Thresholds Artificial Neural Networks
Date of Defense	2006-03-17
Availability	unrestricted
Abstract The Florida Department of Transportation (FDOT) collects vehicle classification data for transportation policy and system planning, traffic operational analysis, safety and accident analysis, and roadway maintenance. FDOT utilizes Scheme F method of classification, which classifies vehicles into 13 vehicle classes according to the number of axles the vehicle has and the lengths between the axles. The vehicle data are collected by inductive loops and piezoelectric sensors installed at more than 300 sites on the state highway system. The Federal Highway Administration (FHWA) requires states’ departments of transportation to report vehicle classification data using Scheme F regardless of the method used in data collection. Moreover, the current FHWA’s Traffic Monitoring Guide allows the states to collect vehicle classification data in urban areas based on the overall vehicle length. The guide states that three or four general vehicle length categories are sufficient for many practical analyses. The guide also provides flexibility for states to select data collection equipments that meet their local and federal traffic data needs and priorities without hindrance from budgets, geographic and organizational constraints. The objective of this research was to develop a length based vehicle classification scheme for Florida. The scheme will be used by non-intrusive traffic detection systems to collect vehicle class data. The task of developing the scheme comprised of collection of vehicle length data throughout the state highway system. The identification of length patterns from the vehicle length data was done using support vector machines. The analysis of the vehicle lengths collected from the Florida state highway system showed three patterns of vehicles: passenger vehicles, single unit trucks and multi-unit trucks. These groups corresponds to classes 1-3, classes 4-7 and classes 8-13 of Scheme F, respectively. The three vehicle categories were defined using length thresholds of 0-21.4 ft, 21.5-42.4 ft, and 42.5 ft and above with an accuracy of 91.1% on the sample data and at least 90.8% on the validation data. The study showed that a large part of misclassification errors was caused by the presence of vehicles towing light trailers. The study recommends the use of an additional variable such as the vehicle profile in order to reduce misclassifications.
Files	Filename       Size       Approximate Download Time (Hours:Minutes:Seconds)     28.8 Modem   56K Modem   ISDN (64 Kb)   ISDN (128 Kb)   Higher-speed Access    TIMTHESIS2006.pdf 694.83 Kb 00:03:13 00:01:39 00:01:26 00:00:43 00:00:03