Typical video surveillance systems are very demanding in terms of infrastructure required to deploy them and also in terms of human resources required to operate them on a continuous basis. In this paper, we propose a system architecture that aims at addressing both issues. The system is composed of multiple intelligent nodes that acquire, process, and archive data/video at a remote site and then automatically generate either alerts or summary reports that are sent to a station at the central operations office of the customer. The intelligent nodes are capable of analyzing multiple types of input data, including video, and take actions ranging from communicating alerts back to the human operator to automatic shutdown of a complete facility. These intelligent nodes serve as the middleware devices in this distributed architecture. We present a case study in the pipelining industry in Canada.