The automated surveillance of persons of interest across multiple videos is considered to be a challenging problem. Person re-identification and human tracking, form the core of the automated surveillance systems. Research has heavily relied on using visual appearance for identification and made many advances in the new feature design and identification accuracy. Meanwhile, more practical problems from the real situation are raised to push researchers to solve them and make the automated surveillance systems more practical. This paper provides a detailed technical review of some well-known successful methods and discusses new ideas to assist re-identification task in solving issues from the real situation such as appearance missing due to occlusion.
展开▼
