Visual Person Searches for Retail Loss Detection: Application and Evaluation

We describe a novel computer-vision based system for facilitating the search for people across multiple non-overlapping cameras. The system has been applied in a retail environment for a variety of problems, most specifically for returns fraud prevention. The system detects and tracks people in multiple cameras and enables rapid cross-camera association of tracks. We have created a human-centred application wherein machine-detected events are browsed and associated in a web-based user interface by a loss-prevention specialist. The system has been tested in a real store environment and we develop a variety of performance measures for the task and present results with a breakdown of error types.