Object Matching in Distributed Video Surveillance Systems by LDA-Based Appearance Descriptors