The security industry has come a long way since the days of going to a video tape archive, pulling the tape, popping it into a VCR, and fast forwarding until you find something. Current technology enables operators to manually or automatically flag an incident based on pre-established criteria; and to then intelligently search based on that criteria or on the proverbial time stamp. Much of this intelligent search is based on person or item characteristics and may employ increasingly sophisticated analytics.
The industry is getting smarter in realizing that the information captured on video equipment may be useful in non-traditional security or non-security applications, including retail assessments, crowd patterns, patient care and monitoring, employee behavior, operations assessment, litigation mitigation and more. Assuming any of these are conducted within legally accepted boundaries, the net effect is to deliver enhanced business-related ROI from the security department. Of course, this presumes that the information useful and relevant, and that security people and their IT and business counterparts are able to work together to connect the dots.
I’ve found several products that can make this process more efficient. My disclaimer is that I have not tested these, so I’m taking the manufacturers claims as stated or as seen in brief demos.
One proven technology, well on its way to becoming mainstream, is pure analytic-based search. Software such as Agent Vi’s Vi-Search, Axxon’s Forensic Search, 3VR’s Video Intelligence Platform (VIP), and IntelliVision’s IntelliSearch, analyzes video streams, generates metadata describing the scene content, and allows for later retrieval and analysis of the video through automated searchwithin the stored metadata. Effectively, all incoming content is indexed and cataloged, enabling efficient recall. From a security standpoint, this approach is well-suited to license plate recognition, facial recognition (within limits), and object and people tracking. Marketing and customer service applications include queue length, dwell time and customer presence.
Still, fast-forwarding is prevalent — either because intelligent search capability is not part of the system, or because the search criteria did not meet the full requirements of the video review. The problem is that operators may miss fast-occurring activity that happens in dropped frames.
A different approach uses basic scene analysis to separate what is moving from what is not (background) and to display only the motion parts. Then, only the dynamic portions of the video are strung together to create the summary. This is the approach of both IntelliVision’s Video Summary and BriefCam’s Video Synopsis. Both products are available for either standalone platform or OEM integration
BriefCam applies further intelligence through the criteria of defining events. When an event is detected by the software, an event is created along with a time date stamp. The software then has a clever way of rendering the video (what BriefCam terms “synopsis”) that allows a rapid view of events by the simultaneous presentation of multiple objects and activities that have occurred at different times. In essence, it packs the motion into as efficient a presentation as possible, without the time constraint of making events sequential.
While this would probably not be effective in applications where there is dense continuous activity or when maintaining chronological order of the edited video is critical, the usual effect is to allow related occurrences to be rapidly searched and subsequently correlated with the original video. Selecting an event from the visual event stream allows the operator to bring up the original video. On average, one hour of video is “synopsized” into roughly a minute.