Assessment of image processing techniques as a means of improving Personal Security in Public Transport

This paper presents work that is currently being developed in the Vision and Robotics Laboratory, King’s College London concerning the improvement of personal security in public railway transport systems. The application environment is focused on London Underground installations and video sequences have been obtained from standard video recordings. In order to achieve a good segmentation and tracking system working in real time with the specified images, we propose a simple segmentation technique based on the following assumptions: the background does not change; people usually wear dark clothes; colour information is not significant; the perspective of the camera usually results in large and fast changes in the size of the people and the use of low spatial resolution improves segmentation results. These assumptions do not excessively constraint the range of application or reduce performance. On the contrary, analysing reduced-resolution-B&W images without updating background allows real-time processing on incoming video images.