A Novel General Purpose GPU Computing Scheme for Automated Multi-Object Surveillance

Akriti Sahu; Rajesh Tiwari

doi:10.37591/rtpc.v2i1.288

Abstract

In this study we presented a Graphic Processing Unit (GPU) based computing scheme for the division of large computational jobs such as of computer vision for the minimal processing time to facilitate the tracking of multiple objects in real time. The system comprises of the two objectives i.e., to dynamically parallelize the jobs and to reduce the CPU (Central processing unit) to GPU communication time. This will facilitate the other non-linear computing jobs to be logically broken into the parts based on the current memory usage criterion of the associated hardware devices. These features allow the robust computational environment for multi-level tasks like that of multiple visions, pattern matching and characterizations.Keywords: Parallel computing, visual surveillance, GPU, multi-core, CPUCite this ArticleAkriti Sahu, Rajesh Tiwari, A Novel General Purpose GPU Computing Scheme for Automated Multi-Object Surveillance, Recent Trends in Parallel Computing. 2015; 2(1): 1–5p.