5.1 Introduction

The number of security and traffic cameras installed in both private and public areas is increasing. Since human guards can only effectively deal with a limited number of monitors and their performance falls as a result of boredom or fatigue, automatic analysis of the video content is required. Examples of promising applications [ ^[1]] are monitoring metro stations [ ^[2]] or detecting highways traffic jams, semantic content retrieval, detection of loitering and ubiquitous surveillance. The requirements for these systems are to be multi-camera, highly reliable and robust and user-friendly. We present a generic, flexible and robust approach for an intelligent distributed real-time video surveillance system. The application detects events of interest in visual scenes, highlights alarms and computes statistics. The three main technical axes are the hardware and system issues, the architecture and middleware and the computer vision functionalities.

The chapter is organised as follows: Section 5.2 describes the requirements of the overall system from the user point of view. Section 5.3 presents a global description of the system and its main hardware characteristics. Section 5.4 illustrates how the hardware and software in a distributed system are integrated via the middleware. Section 5.5 goes deeper into an understanding of the image analysis module, which produces metadata information. Section 5.6 is devoted to the description of a case study, which is that of counting people. Some references to this case study are made all along...