Multi-View Person Re-IDentification with Wireless Multimedia Sensor Networks

Wireless Multimedia Sensor Networks (WMSN) involves a set of multimedia sensors such as cameras to collect video or audio data about the deployed Region of Interest (RoI). Compared to traditional Wireless Sensor Networks, the visible information of covered targets can be obtained with WMSN. Moreover, due to dense deployment of low-cost cameras, the object such as person may be covered by several cameras. This paper introduces a new multi-view person Re-IDentification (Re-ID) model with WMSN. Moreover, human skeleton extraction based person Re-ID method is applied to extract person’s posture feature. As a result, significant cost by labeling data will be avoided. Finally, Decision-level and Feature-level feature fusion methods with multi-view cameras are proposed to complete Re-Identification process. Multi-view video-based than Re-ID algorithm has better performance than single-view video-based Re-ID algorithm. As we know, this is to first research on person Re-Identification with WMSN.