Multimedia data model for video-detection research

Traffic monitoring and detection systems are important for constructing closed-loop traffic-signal control, freeway-control, and congestion-management systems. Most detection and monitoring systems depend on inductive loops. Recently, many agencies have been investigating video-image–based detection systems because cameras can collect data on several traffic lanes and can be installed quite rapidly on an adjacent pole or bridge. User consensus is that the basic hardware architecture of these systems is well defined. However, there is room for significant improvements in performance and capability of video-detection algorithms. Several researchers are currently working to develop new algorithms. However, due to technology limitations these algorithms have traditionally been evaluated at a macroscopic level by comparing counts obtained by loop detectors with the image-based detection system. This paper discusses the development of a multimedia database-management system for investigating the microscopic performance of video-detection algorithms. The paper describes the process of acquiring the proper data, a definition for a suitable database schema, and the peripheral database clients required to evaluate detection algorithms.