Since the launch of Google’s search engine in 1998, Internet search has grown and expanded from desktops to mobile devices, generating over 3.5 billion searches of text, images and videos per day. One likely reason for this popularity is the search engine’s ability to immediately produce relevant results based on search criteria.
The security industry, in contrast, has not always taken advantage of such innovations in search. The right search technology exists to help loss prevention and fraud investigators find security video evidence of criminal activity. Yet today, most investigators are watching endless hours of video instead of leveraging specific criteria to pinpoint the moment an event of interest occurs. Why can’t security video search perform like Google?
Consider a large bank with locations around the globe and 150,000 cameras monitoring these locations. With just one hour of recorded video per day per camera, the bank is generating over 50 million hours of video per year. Yet the fraud team staff is comprised of only 10 full time reviewers. With 2000 work hours per year, this team is only capable of reviewing 20,000 hours of video per year, or just .04 percent of what is being captured.
This example illustrates a common problem in security today. Most investigators conduct linear timeline searches, meaning they either watch live feeds of specific cameras or recorded video from a specific day when an event occurred. This often requires downloading the entire day’s video from a central storage server and then watching countless hours of video to find events of interest. As in the global bank example, there is simply too much video to watch for this approach to be successful. Not to mention the toll this can take on the corporate network, with bandwidth heavy video downloads interfering with critical application performance.
Although video is the fastest growing type of big data today, it is more complicated to search video compared to text-based data. Keywords and descriptions of videos must be either automated or entered manually in contrast to text-based data that inherently includes keywords. Visual data such as video is complex, with layers of information and context about how people behave and interact. Software can deliver an unbiased analysis of this behavior whereas humans will naturally insert bias when searching and viewing video. All of these factors make it more challenging to efficiently search video footage.
Today’s explosive growth in the amount of video data is driving demand for more effective and efficient video search. Recent advances in video management systems (VMS) are now making it possible to leverage surveillance video data for both loss and fraud prevention, as well customer insights, such as in-store buying habits and trends. Leading VMS solutions include the basic store and record functionality plus a plug-in framework for analytics, enterprise search and case management, all in one system. A more effective search capability allows investigators to quickly search vast volumes of video and find relevant evidence in minutes by using specific criteria such as facial surveillance, license plate recognition and object tracking analytics.
Many VMS products also include context integrations that can correlate video footage with Access Control, ATM or POS transaction data. Leading businesses are now taking advantage of this same technology to analyze video data to better understand customer behavior and operational metrics. Analytics such as people counting, dwell patterns, demographics and directional heat maps can be used to improve the customer experience, optimize staffing levels, increase conversion rates and analyze performance across locations.
In one example, a commercial parking firm had been relying on cameras recording the number of cars parked in their facilities to ensure this matched the tickets issued at the entry gate. Even with this video recorded, employees were still required to walk the garages every night to manually record every car by color and license plate to accurately charge those cars that were left overnight.
Once the facility switched to a newer VMS that included a more efficient search capability and license plate recognition analytic, the system was able to automatically capture all license plates that entered the garage. So if a customer loses their ticket, they can simply provide their license plate number so customer service reps can run a search and verify how long the car has been in the lot. This enables the company to accurately charge customers, preserve revenue, and save countless employee hours of walking the garages every night.
This type of forensic search capability has also helped to reduce employee theft and shoplifting investigation times for a major retailer. The company had a team of five loss prevention managers working at maximum capacity to investigate activity at over 1200 stores. By implementing a more effective video search solution, they were able to reduce investigation times from one week to 20 minutes and reduce shrink by 60%. The newer VMS solutions are also so cost effective that this company achieved a return on investment in less than 10 months.
In addition to improvements in security, businesses can leverage this same video data to learn about customer behavior, improve operations and enhance other important data sources such as POS or ATM and teller transaction data. Retail stores are leveraging video analytics to more accurately measure in-store traffic, analyze associated traffic patterns (including demographics) and conversion rates to improve product placement and compare store performance across locations. Queue management analytics can help businesses improve store operations with more efficient staff planning. Merchandising management can also benefit from a better understanding of accurate conversions based upon day and time to manage inventories and product placement.
For businesses that are ready to consider an upgrade to their security and surveillance system, and take advantage of these new capabilities to improve loss investigations and better understand customer behavior, be sure to evaluate different solutions using the following VMS checklist:
• Handle the basics of store and record
• Deliver LP and BI leveraging one platform
• Enterprise scale and capabilities
• Analytics and context integration platform
• Advanced search architecture
• Multiple deployment options (hardware, software, on-premise, cloud)
• Easily configured and administered
By choosing a VMS that meets these criteria, businesses can experience video search that parallels the Google experience.
By Al Shipp for SourceSecurity.com