The implementation of a vision sensor for traffic surveillance

The EC project OMNI ("Open Model For Network-wide Heterogeneous Intersection-based Transport Management") [1] provides an open-architecture model for road traffic management, by defining: standard interfacing for new applications or devices, and the integration of technology (surveillance applications, advanced sensors, other urban traffic strategies, etc) from different vendors. One of the devices, specified in the OMNI model, is the video sensor [2], [3]. In this paper we describe our design and implementation of a modular video sensor software. We have implemented following vision modules within OMNI [4]: (1) camera auto-calibration, (2) traffic flow and (3) queue length detection, (4) car plate recognition. An additional communication module provides an interface to the OMNI-MOUN interface. A programming class implements each vision module and it can be transformed into an independent process or object, if a distributed system is required for it, by using the libraries of DCOM or CORBA.