A Real-Time Smart Assistant for Video Surveillance Through Handheld Devices

Abstract

In a remote surveillance system, a high resolution surveillance camera streams its video to a user's handheld device. Such devices are unable to make use of the high resolution video due to their limited display size and bandwidth. In this paper, we propose a method to assist the mobile operator of the surveillance camera in focusing on sensitive regions of the video. Our system automatically identifies relevant regions. We introduce a pan and zoom strategy to ensure that the operator is able to see fine details in these areas while maintaining contextual knowledge. Regions of interest are identified using foreground detection as well as face and body detection. The efficacy of the proposed method is demonstrated through a user study. Our proposed method was reported to be more useful than two comparable approaches for getting an understanding of the activities in a surveillance scene while maintaining context.

sZoom: A Framework for Automatic Zoom into High Resolution Surveillance Videos

“…The goal of this experiment is to reduce the processing time without affecting the accuracy significantly. The third experiment compares the accuracy and the timing performance of the proposed system with the state of the art [2]. Lastly, we assess the effectiveness of the threat model, coverage, and overall zoom operation through three user studies.…”

“…Hence, we get an improved prediction at the cost of small latency. In the previous work [2], the latency was a full zoom window, i.e., ∆ frames. We are able to reduce it to ω frames because of the ROI tracking and corresponding iterative refinement of the target ROI during the zoom operation (Section II-C).…”

“…With the incorporation of tracking and refinements in the framework, we are able to improve the performance of sZoom compared to the preliminary version of the work [2]. The improvement is two-fold: reduced processing time and improved accuracy.…”

“…The third experiment compares the proposed work with the preliminary version of the work that was published at as a short paper [2]. In this paper we improve the previous work with better sensitivity detection process, reduced latency, and better coverage of moving targets.…”

“…The proposed method improves the amount of information gained by security operators by 99%. • We propose optimizations in the system that improve the accuracy by 42% and reduce the processing time to 5% of the original time making the sZoomed video more suitable for surveillance with 36% higher ratings in comparison to the previous work [2]. The rest of the paper is organized as follows.…”