Tracking people and authenticating their activities are important parts of the element technology required to establish a system that will automatically detect suspicious people using monitoring cameras installed in station. The technology must also transmit this information for analysis to control centers, monitoring centers and other relevant organizations, and provide those concerned with appropriate instructions. An algorithm to predict movement using Karman filter was therefore developed which can track objects even in a crowd. A verification test with an image of several dozen pedestrians demonstrated that this algorithm was able to accurately track people who are crossing or hiding behind obstacles (Fig. 1). Analyzing a scene of violence and fallen person, it was confirmed that the system was able to detect these abnormal events by capturing unnatural human posture and limb movement (Fig. 2).

Two systems were also developed. The first one grasps location of station crew indoor and outdoor through both the use of active wireless tags (indoor) and GPS (outdoor).The second one selects automatically appropriate recipients (control room, station office, station crew the closest to abnormal event, security companies etc.) according to type of abnormal event, degree of urgency and location of station crew (Fig. 3). It is confirmed through laboratory tests that these systems function correctly.

This research was executed with a subsidy from the Ministry of Land, Infrastructure and Transport.