A real-time networked camera system:a scheduled distributed camera system reduces the latency

This report presents the results of a Real-time Networked Camera System, com-missioned by the SAN Group in TU/e. Distributed Systems are motivated by two reasons, the first reason is the physical environment as a requirement and the second reason is to provide a better Quality of Service (QoS). This project describes the distributed system with a video processing application. The aim is to deal with the distributed system as one system thus minimizing delays while keeping the predictability in a real-time context. Time is the most crucial ingredient for the real-time systems in the sense that the tasks within the application should meet with the task deadline. With respect to the distributed system we need to consider a couple of issues. The first one is to have a distributed system and a modular application that is mapped to multiple system nodes. The second issue is to schedule the modules collectively and the third is to propose a solution when shared resource(s) (such as the network) are required by several nodes at the same time. In order to provide a distributed system, we connect 2 cameras with 1 PC via a network switch. Video processing has two parts; the first part consists of creating a frame, encoding the frame, and streaming it to the network and the second part deals with receiving the frame, decoding the frame, and displaying the frame. The first part is running on the cameras and the second part is running on the PC. In order to give real-time behavior to the system, the system components should provide the real-time behavior. The camera is installed with the µC/OS-II (Open Source Real-time Kernel). We investigated the Real-time Operating System and its installation on the PC. In order to provide resource management to the shared resources, we designed and implemented Admission control which controls access to the required con-nection to the PC. We designed and implemented a component to delay the start of any of the cameras in order to synchronize the network utilization. We also designed an enforcement component to allow the tasks to run as much as they should and monitor the frames streamed to the network. The results show that with the Admission Control, cameras only send as many frames as the network can transport. The given start delay to the system shows that overlap can be prevented, but we could not evaluate it because of the semi-tested/unreleased code which is provided by the camera providers. The source code we used is the test source code which was not mature