Bill Siuru, Ph.D, PE
Video analytics, also called video content analysis, is needed today as the availability of digital recording devices explodes exponentially.
Surveillance cameras monitoring a middle school detect a suspicious person lurking on the school grounds. This triggers an alert sent directly to the local police, even before the person tries to enter the school building and the alarm system is activated. Thus, the police can be dispatched to investigate a few minutes sooner, perhaps fast enough to avert another school shooting. What is key here is the alerting was done without a human viewing a CCTV monitor, but done using video analytics technology.
High-quality video surveillance cameras are now mounted everywhere: government buildings, street intersections, schools, private businesses, and homes. They are supplemented by “self-appointed videographers” with smartphones which take live streaming video of anything they find remotely interesting. Then, there are dash cams in police vehicles and the majority of police officers now wear a body cam.
It is impossible for even a large staff of humans to monitor all this video data 24/7/365. Even if they could, watching surveillance video is tiring, making it difficult to focus and identify threats as they develop. It is estimated that, after 20 minutes of watching surveillance footage, reviewers miss 95% percent of incidents.
Proactive Video Analytics
Video analytics software can monitor multiple video feeds around the clock detecting any criminal or unusual activity in progress and alerting human monitors only when it detects something out of the ordinary. This allows law enforcement resources to respond faster than when 911 calls and dispatchers are involved. With any luck, it detects crimes which are in progress, increasing the chances of catching the bad guys.
For example, among hundreds of video feeds coming into a police crime center, one camera detects activity which is not normal via analytics software. Long after business hours, a building-mounted camera observes a car with its headlights turned off park next to a building. As the person approaches the side door of the building, the software sends an alert to a crime center analyst. Sophisticated software using computer-based artificial intelligent analysis has detected the anomalous behavior from previously learned behaviors using machine learning. The system generated alert was accompanied by a short video clip of the events leading up to the alert. The analyst quickly concludes that they are watching a burglary in progress and dispatches police units. Within minutes, officers surround the store and arrest the person.
Catching the Perp and Solving Crimes
In the past, LEO investigators looking for a person of interest might have to review hundreds of hours of video from multiple cameras. Video analytics permit focusing only on people who fit a particular description. Once someone has been identified as a person of interest, the system can locate that same person on other video feeds, even if they span many days and locations. Deep learning technologies incorporated in video analytics systems can find details which previously required human eyes. They can distinguish men from women, children from adults, etc., based on size, gait, walking speed, and other subtle features. Filters can find people dressed in red clothing from those in other colors or find just those with backpacks.
If the appearance, gender and clothing of the suspect is known, these features can be used as filters to limit the search. Investigators then are shown only footage which matches the criteria. If a digital photo of the suspect is available and the analytics software incorporates facial recognition, this becomes an additional filter. As an example, if the subject drove a silver SUV, this criteria could be entered into the filtering software and only footage with silver SUVs would be shown to the human reviewer.
Stakeouts which now require teams to watch often for days at a time may no longer be needed. This longtime police job can be done automated by a covert surveillance camera(s) and video analytics doing the watching.
Surveillance cameras and video analytics could help police locate seniors with dementia who have wandered away. They can identify pedestrians who are moving unsteadily or slower than normal and then alert officers to perform a welfare check, even before a caregivers realizes the senior is missing. A video analytics system can be shown a missing child’s photo, then it can find him (or her) on every video feed which is available.
Making Best Use of Resources
The cost of surveillance systems and video analytics technology can be substantial and, thus, a police department can find funding difficult. However, over the long haul, it’s usually much less expensive than a room of human monitors, even if they are nonsworn personnel. With video analytics, hours of surveillance video needed to be reviewed turn into minutes, or even, seconds. Video analytics technology can be a tremendous force multiplier for law enforcement agencies struggling with staffing and can maximize and leverage existing resources.
As might be expected, privacy concerns could impede implementation of video analytics systems. Civil libertarians are already opposing the use of this technology. San Francisco is the first US city to ban the use of facial recognition software, an important tool in video analytics, by police and other city departments.
Bill Siuru is a retired USAF colonel. He has a Ph.D. in mechanical engineering from Arizona State University. He has been writing about automotive, aviation and technology subjects for many years.
EXAMPLES OF AVAILABLE VIDEO ANALYTICS TECHNOLOGY
BriefCam®, a leading provider of Video Content Analytics and VIDEO SYNOPSIS® solutions, now offers version v5.4 of its video content analytics platform. This version advances real-time capabilities, enhances user experience and accelerates processing performance across all three of the platform’s three modules. BriefCam REVIEW’s multicamera search can identify men, women, children, vehicles, and lighting changes. It offers 27 classes and attributes, in addition to face recognition, appearance similarity, color, size, speed, path, direction, and dwell time. BriefCam RESPOND provides real-time alerts of wanted suspects turning up on camera when their facial recognition data is input. BriefCam RESEARCH creates and customizes interactive, intuitive dashboards with chart suggestions which autogenerate reports and Insight Advisor which prioritizes relevant data points. It can be used to map out crime patterns within a given locale over time for improved policing strategies. (briefcam.com)
Kinesense is a computer vision and video analytics company based in Dublin, Ireland. The company’s computer vision products use a combination of motion detection and deep learning algorithms technology to search CCTV content during criminal investigations. (kinesense-vca.com)
GenVis, an Australian-based visual artificial intelligence company, unveiled their new video analytics solution, MOTHERSHIP at ResponderXLive ’18. Its products have been developed by working closely with law enforcement agencies in Australia and are aimed at finding suspects in real time or though post event video analysis. MOTHERSHIP by GenVis uses proprietary AI and facial recognition to analyze video files and live camera feeds. GenVis is now working with LE agencies in the US and Canada. (genvis.co)
IntelliVision Video Analytics’ video analytics automatically analyze live streaming video from CCTV and surveillance cameras to provide actionable alerts for security events, eliminating false alarms and reducing the need for manual monitoring. Video Analytics extract only the valid motion in a scene, filtering out noise such as lighting changes, weather, trees and animal movements. It can detect, for example, any attempt to tamper with the camera, a moving object which crosses a defined line, if an object has been left unattended for too long or when a specific object has been removed from the scene. It can also detect a loitering person, someone who has fallen or is under duress, or a person running. (intelli-vision.com)
Agent Video Intelligence (Agent Vi). Israeli security provider Agent Vi offers innoVi, a comprehensive video analysis solution which provides real-time warnings, combined with a high-speed search engine. The cloud-based software includes an advanced algorithm which makes it possible to detect changes in video content streamed to it, and to identify and classify what it detects in a clip. innoVI is based on cloud computing technology which allows access from any location. (agentvi.com)