Overview
Our system, illustrated right, interfaces a video surveillance system with our labs computational network. Data is pulled in for analysis on a schedule, via the Acquisition Module. Data may be gathered directly from camera devices or from Network Video Recorders (NVR). An extra admin layer to catalog system attributes and maintain timing across devices is provided by the System Manager (SM).

This module accepts user input to specify which data to retrieve from the surveillance network. This could, for example, specify (i) from which camera we want data, (ii) the pan tilt zoom (PTZ) of the desired camera, or (iii) whether we want live or past-recorded data. These specifications are parsed and interpreted in our data binding and web service scheduler functions. Lastly, the desired web services are passed to the surveillance network via the Pelco API Library. Video data is then streamed and stacked to a buffer, awaiting processing by the Analysis Module.

Timing Schema
User Input Schema is divided into two sections: Timing Parameters and Scene Configuration. User input is specified in a text file. Timing Parameters specify the timing schedule illustrated, right. Scene Configuration specifies camera settings (i.e. which camera, PTZ coordinates, codec compression, resolution, etc.). For example: if we request a 20-second clip every 10 minutes and also specify two scenes, we will receive two 20-second clips every 10 minutes; one for each scene. The illustration to the right shows the timing schedule containing two clips. Their start times, respectively, are t=k and t=k+1. Not all clips will require LAG1, e.g. static cameras. This application was designed to cycle through multiple scenes/cameras to gather data from multiple areas of interest. Although Pelco software supports multiple streams, this application cycles through scenes and exports single clips at a time until the lifetime of the experiment expires.