A distributed solution to detect targets in crowds using visual sensor networks

Visual sensor networks (VSNs) have been attracting more and more research attention nowadays. This paper presents our design of a distributed solution for a VSN to detect targets in a large and crowded field. Due to ambiguities in local visual information and ubiquitous visual occlusions, obtaining accurate detection results requires fully utilizing the entire visual information acquired across the network. In our solution, a Back-Projecting-Flooding (BPF) method is proposed to integrate related visual information, which is followed by a Distributed-Highest-Confidence-First (DHCF) method that implements an iterative global target selection process in a distributed manner. Compared with its centralized version, this solution not only achieves equivalent detection accuracy but also reduces and balances the amount of energy consumption across the network. Simulation results are presented to show the effectiveness of this solution.