SSA: Smart Surveillance Assistant for Mobile Devices

Abstract

Over the past few years, the capability of smart devices has grown incessantly, and is showing no sign of slowing down. Along with the decreasing cost of manufacturing high definition surveillance cameras, this has led to the increased ease and convenience of installing surveillance cameras at or in private places. Various applications focus on home surveillance, however simply sending a high definition image does not satisfy all of the users’ needs. On a site that is monitored by a surveillance camera, the high-resolution surveillance camera streams its video to a user's handheld device. Unfortunately, such devices are unable to make use of the high-resolution video due to their limited display size and bandwidth, and the visual range is also a key problem.

In this thesis, we propose a method to assist the mobile operator of the surveillance camera in focusing on sensitive regions of the videos. Our system automatically identifies relevant regions, combined with foreground detection and human body detection, which is referred to as object detection. A sensitivity map that represents those informative regions returned by the detection methods is accumulated over number of frames. It shows the collection of the sensitivity data in the video over a period of time. We then introduce a zoom strategy to ensure that the operator is able to see the fine details in these areas, while maintaining contextual knowledge. Regions of interest are identified using foreground detection as well as body detection. In order to accelerate the processing speed, we propose two optimization methods. The efficacy of the proposed methods is demonstrated through a user study, the results of which show that this approach is more successful than three comparable approaches used to get an understanding of the activities in a surveillance scene while maintaining context.