An Investigation in the Use of Inductive Loop Signatures for Vehicle Classification

The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the State of California. This report does not constitute a standard, specification, or regulation. ABSTRACT This final report describes an advanced traffic surveillance technique based on pattern recognition and the use of current inductive loop technology. The focus of the investigation was a study of the feasibility of using inductive loop signatures for obtaining vehicle classification information on a network-wide level. The potential benefits from the vehicle classification information include improvements in vehicle reidentification algorithms, roadway maintenance, vehicle emissions management, roadway design, traffic modeling and simulation, traffic safety, and automatic toll collection. The compilation of a vehicle classification database is also a valuable resource for researchers in the areas of transportation planning and control. This effort complements current PATH (Partners for Advanced Transit and Highways) research in advanced surveillance technology. Different pattern recognition techniques for vehicle classification such as classical decision theoretic approach and advanced neural networks were employed in this research. Inductive signatures from the SR-24 freeway were used to test vehicle classification algorithms. Classification rates of greater than 80% were obtained using different datasets. The results demonstrated the feasibility and the potential of using this method for collecting vehicle classification data. EXECUTIVE SUMMARY Vehicle classification is the process of separating vehicles according to different predefined classes. Vehicle classification information can be used in many transportation of vehicle classification data has previously been investigated using various new surveillance technologies such as video, laser, and acoustic detectors. One goal of this project is to exploit the current inductive loop infrastructure. However, this fact does not abolish the possibility of changing the vehicle classification algorithms developed in this research to accept other forms of detector output in the future. In fact, detector fusion can the wave of the future for traffic surveillance once the economics of multi-detector surveillance become more reasonable. This final report presents two distinct vehicle classification methods using the following seven vehicle classes: car, SUV/pickup, van, limousine, bus, two axle trucks, and trucks with greater than two axles. These seven vehicle classes are interesting to study because they cannot be determined by using axle counters, and they can be used in different transportation applications. Both classification methods utilize vehicle inductive signatures …