SBIR-STTR Award

The goal of this project will be to build on an existing product that was developed under a Transportation Research Board / Ideas Deserving Exploratory Analysis project. This program produced a product called Autosense II, which provides classification, speed, lane position and towbar detection of a vehicle in a traffic lane at speed up to 100mph. The Autosense II provides highresolution 3-Dimensional profiling, but relies on a relatively simple rule-based alorithm for vehicle classification. Designed to operate on a low-cost microcontroller inside the device, the rule-based algorithm uses only the height, lenght and width for classification. The proposed research project will explore the feasibility of improving the classification by using the Autosense IIÕs 3-D profile data combined with a more sophisticated pattern matching algorthm similar to the ones developed under Department of Defense contracts for submunition scanning laser radar guidance.Anticipated results/potential commercial applicationsIt is anticipated that the results will lead to an overhead sensor capable of automatically classifying not only broad groupings of the vehicle classes (e.g., automobiles, light trucks, single unit trucks, and combination trucks), but also identifying specific body types within the vehicle classes.

When truck studies are undertaken, data is often collected by stopping vehicles and manually recording the vehicle body characteristics and other information. This is a labor-intensive and time-consuming process that could be eliminated through the use of automated classification techniques. Most automated techniques, however, employee owned axle counting as a means of classification, thus significantly limiting the amount of data obtained. At automated system that could provide information on vehicle body shape would add much-needed data to an automated classification system. The results of the Phase 1 research effort to demonstrated the feasibility off automatically classifying vehicles by their body types by using a high-resolution laser radar. By developing an algorithm that combined pattern matching, feature analysis, structural analysis and rule-based logic semi-trailers with various types of cabs and trailers were classified. When running on a PC, the algorithm shown to be capable of classifying vehicles in real-time as they passed under the sensor. The Phase 2 effort involves developing and demonstrated the classification device. Tasks include applications research, algorithm development, processor board development, and mobile test system development testing. Anticipated results and potential commercial applications of results: It is anticipated that the results lead to an overhead sensor capable of automatically classifying not only broad groupings of the vehicle classes (e.g. automobiles, light trucks, single unit trucks, and combination trucks) but also identifying specific body types within the vehicle classes. In addition, truck classification performance in applications will be demonstrated.

Keywords:
Vehicle classification; laser radar; pattern matching algorithm