Person re-identification is the task of recognizing or identifying a personacross multiple views in multi-camera networks. Although there has been muchprogress in person re-identification, person re-identification in large-scalemulti-camera networks still remains a challenging task because of the largespatio-temporal uncertainty and high complexity due to a large number ofcameras and people. To handle these difficulties, additional information suchas camera network topology should be provided, which is also difficult toautomatically estimate, unfortunately. In this study, we propose a unifiedframework which jointly solves both person re-identification and camera networktopology inference problems with minimal prior knowledge about theenvironments. The proposed framework takes general multi-camera networkenvironments into account and can be applied to online person re-identificationin large-scale multi-camera networks. In addition, to effectively show thesuperiority of the proposed framework, we provide a new personre-identification dataset with full annotations, named SLP, captured in themulti-camera network consisting of nine non-overlapping cameras. Experimentalresults using our person re-identification and public datasets show that theproposed methods are promising for both person re-identification and cameratopology inference tasks.
展开▼
