With the increasing availability of cheap imaging sensors and ubiquitous computer networking, large networks of cameras are being installed in many transportation environments, such as freeway systems, business structures and, and metropolitan roadways. However, most applications of computer vision to Intelligent Transport Systems have dealt with data from single sensors. Even systems with multiple cameras typically process the data from each sensor independently. Data fusion among sensors in a network allows for applications impossible with standalone sensors and processing systems. These include such applications as wide-area tracking and the calculation of trip times for individual vehicles. This paper describes a distributed computing system capable of managing an arbitrarily large sensor network using common computing and networking platforms. The architecture is capable of handling most common computer vision tasks, as well as the inter-sensor communication necessary for developing new algorithms, which employ data from multiple sensors. This system is tested with a wide-area tracking algorithm which tracks moving objects through a wide-area camera network with non-overlapping fields of view. The sensor network covers a large area of a college campus, and an adjacent interstate freeway. This algorithm allows the system to maintain the identity of tracked objects as they leave and enter the fields of view of individual sensors. Such an algorithm is necessary for applications, which require tracking objects over large distances or over long periods of time in an environment without complete sensor coverage.
Samenvatting