Real-time Person Tracking and Re-identification Based on Feature Matching

Abstract

The existing video-oriented person tracking and re- recognition research are all based on face recognition, which cannot meet the requirements of high continuity and high accuracy. In order to solve this problem, this paper propose a tracking and re-identification scheme based on KM (Kuhn- Munkras) algorithm. When tracking a person, this scheme establishes the connection between the face image observed in real time, the face database, and the human body image observed in real time to determine their identity. During tracking person, the appearance characteristics can be captured all the time, and the recognition of person has high real-time performance and high continuity. In the tracking process, a real-time updated person information cache list is established to realize the re- identification function after the person tracking is interrupted for a long time.