|
Video Analytics Beginning to Deliver on Promises
By Nicholas Samanich
Published on: May 22, 2008
Industry: Medium & Large Business: Retail, Banking & Financial, General Business
Solution: Smart EAS, Commercial Solutions
Analytic software uses algorithms that help identify security vulnerabilities and detect behaviors (and actions) that can lead to security events. This whitepaper describes the evolution of analytics over the past few years and how they can be applied in a variety of ways to proactively improve security solutions, business processes and productivity. Image is everything – marketing people have been telling us that for years and a popular maker of consumer cameras has even used it as an advertising slogan. It may be a simplistic overstatement, but for the security industry right now image, if not everything, is “the thing.” How we capture and analyze video images and what these images can ultimately tell us is rapidly changing the industry and the scope of services provided. We are at the beginning of a dramatic transition in video technology and applications with video analytics, sometimes referred to as video content analysis, as a key driver.
Video analytics bring sophisticated information technology to security solutions and move typical systems beyond basic security. They not only provide information that can help secure people and property, but also improve business processes and productivity. Part of what has been called convergence in the industry is the blending of IP (Internet protocol) enabled analytic software and firmware with typical security applications such as access control and video monitoring. This merging of technologies is producing a powerful set of business tools that take security solutions to new and higher levels.
Traditionally, electronic security has been reactive, meaning the system waits for an anomaly to occur. An event trips an alarm and the system reacts. Analytics “see” much more than the standard system and do more than just react to anomalies. They use algorithms that can anticipate and predict actions and can help identify security vulnerabilities in the environment. This makes for a proactive system that may prevent certain actions or, at the very least, allow for a quicker and more effective reaction. The data gathered from analytics also provides insight that can lead to proactive design improvements in security solutions. With analytics, security professionals can more quickly identify problem areas or trouble spots and work to build in measures to secure these areas rather than just wait to react to events.
The evolution of analytics has been dramatic over the last few years as they have become increasingly sophisticated. However, there have been a number of technical challenges that have stunted commercial adoption. Some analytics work best when they are applied at the “edge”. A primary example is the use of an analytic to filter the video content that gets sent on through the network. This is inconsistent with server based analytic solutions which have been the standard up to now. Providers have begun to push analytics closer to the edge by embedding them in field appliances and now in cameras. Often the challenge in pushing video to the edge is that there is significant processing that must occur there including pushing raw video through analytic algorithms, compressing video, and transmitting video. Several providers are now pursuing a decoupled approach where they are providing a solution that only processes the analytic and outputs the result as metadata. Video processing and transport is left to the video management solution. In addition, new custom and enhanced industry standard Digital Signal Processor (DSP) chips are now being introduced that will deliver more processing capacity. This will have dual benefit in that they will enable more edge processing while significantly reducing the cost of analytic processing by eliminating or downsizing server requirements.
In addition, analytics continue to be sensitive to environmental conditions such as changing light and shade patterns. As more analytics solutions are deployed, lessons learned are being captured and fed back into both the design of the algorithms and the deployment standards thus making this less of an art and more of a science.
 | In this point-of-sale system screen capture, the employee is processing a return trasaction even though there is no customer detected at the counter. The intelligent video generates an exception alert, which is included in a report for the loss prevention manager. Image courtesy ObjectVideo. |
Currently, some applications are more consistent and reliable than others. The more highly refined and proven technologies include:
Perimeter Protection - A number of companies offer software that can help detect objects entering a predefined area. The software creates a virtual electronic tripwire. When an object crosses this virtual line, the software starts recording input from cameras focused on the area.
Left Object - Analytic software can accurately identify new static changes to the video field. For example, it can alert operators to a briefcase left unattended in an airport terminal or new graffiti spray painted on a building.
Automated Number Plate Recognition (ANPR) - This technology captures vehicle license plate numbers and automatically searches an extensive database to match them with registration information. Law enforcement officials can quickly retrieve vehicle ownership information directly from the analytic software. It also can track movement of assets marked with a number plate such as shipping containers.
Line (Queue) Analysis - Retailers know that long checkout lines turn away customers. Software that analyzes checkout lines is useful in helping plan staffing levels and in providing better customer service. This technology can also provide a long-queue alert in real time so the store manager can redeploy staff to the registers.
People Counting - Aptly named, this software counts people as they enter a business or area. It provides businesses with valuable information on traffic patterns and timing. Also, when combined with point of sale (POS) data, people counting can calculate conversion rates (i.e., sales per customer entering store), which is a key metric for retailers. As people counting is refined to the department level, it will provide even more specific and detailed data.
POS Exception Analysis - Retailers have been analyzing point of sale information for a number of years, but the new video analytics provide intelligent monitoring that can recognize activities such as a void or cash return transaction with no one in line. The event is automatically recorded, described and stored.
Other, more complex analytics are still being refined and perfected including:
Facial Recognition - Software that recognizes facial structures as an authentication mechanism in cooperative access control situations where an employee allows a match of his or her face with an existing photo. While facial recognition can be useful in non-cooperative or passive situations, such as scanning individuals passing through choke points, it is still being fully developed and polished.
Behavior Anomalies - Software that looks for specific behavior that is out of the norm is now being developed. For example, software is being developed to tell when someone is running, fighting or acting overly agitated.
Flow Analysis - This tool learns typical flow patterns in a busy store, airport or other public area. It will recognize when the flow pattern is disrupted. For instance, it would record and store an event where someone is walking in the opposite direction of the traffic flow.
These are just some of the available video analytics tools. Many more have been and are being developed. The value of these technologies is not just that the software learns to analyze situations and record them, but it also describes the captured video and then distributes these descriptions (called metadata) for storage and analysis. In perimeter protection, for example, the software “sees” something cross the tripwire, records and compresses the video and labels it with a description such as “red car enters perimeter.” Using a protected Web site, these databases can be centralized and easily shared.
The advantages of gathering video using analytics are enormous. By using analytics, only relevant activities are recorded or forwarded. As a result, less video is transferred and stored. That requires less bandwidth and less disk space. And since the storage and handling of large amounts of video can be burdensome, analytics help keep the volume of recorded video to a minimum and can justify maintaining only “video of interest.”
Another major benefit of analytics is more efficient forensic processing. The retrieval of relevant video becomes much easier. Finding the video clips of the red car or of similar cars passing through the tripwire over an extended period of time no longer takes a security or law enforcement professional hours to sift through recorded video. With analytics, the video is already labeled in a data file and can be retrieved with ease. This is a huge time saver and not only can result in a quicker reaction to an event, but also allows security professional to spend less time on mundane tasks.
Analytics are at the forefront of a dramatic transition in video monitoring and in the security industry in general. The capabilities of analytics are rapidly changing. Over the next few years, changes will come even more quickly. New algorithms will produce more sophisticated software capable of learning patterns and distinguishing behaviors. Prices will fall as algorithms become embedded in chip-based solutions. The sharing of video will become easier. Using the Internet, video will be retrieved and analyzed from anywhere. Over the next three years, video will be used increasingly for applications other than security such as for operations and management purposes.
Businesses will outsource data analysis and storage to take advantage of highly specialized analytics on demand without incurring the total cost. Complex post analytics will be completed on existing video for specific analysis. All of this means that analytics will continue to push the explosion in video and the information we can gather. Video solutions will be smarter and more efficient. The business of security will look very different from five years ago. It will be based on business applications and will allow for more effective security solutions that are specific to each end user’s needs. |
