Real-time processing for intelligent-surveillance applications

Abstract

Although the typical sensor-server model has been widely used in intelligent-surveillance applications, the workload increases for the centralized server as the number of sensors increases; therefore, the provision of a scalable performance requires the distribution of the centralized-server workload into the sensors. Due to the limited resources of the sensor side, however, a resource-efficient real-time processing technique is required. In this paper, a real-time sensor-side surveillance technique for which parallel processing is used on CPU-GPU hybrid-computing devices is proposed. The experiment results reveal that the proposed method can provide a real-time execution of the surveillance system.