Fast and Lightweight Online Person Search for Large-Scale Surveillance Systems

The demand for methods for video analysis in the field of surveillance technology is rapidly growing due to the increasing amount of surveillance footage available. Intelligent methods for surveillance software offer numerous possibilities to support police investigations and crime prevention. This includes the integration of video processing pipelines for tasks such as detection of graffiti, suspicious luggage, or intruders. Another important surveillance task is the semi-automated search for specific persons-of-interest within a camera network. In this work, we identify the major obstacles for the development of person search systems as the real-time processing capability on affordable hardware and the performance gap of person detection and re-identification methods on unseen target domain data. In addition, we demonstrate the current potential of intelligent online person search by developing a real-world, large-scale surveillance system. An extensive evaluation is provided for person detection, tracking, and re-identification components on affordable hardware setups, for which the whole system achieves real-time processing up to 76 FPS.