Page 56 - Security Today, October 2018

Video intelligence can provide critical support for law enforcement teams building a case but allocating the resources to watching and analyzing video is a major challenge. Reviewing video evidence re- quires valuable time and manpower, and, often, investigative teams must decide to rely less on time-consuming video intelligence to better use the available resources. The police team might decide to limit video investigation and only review footage from three cameras focused on a single doorway—when they could derive much more detail and accuracy by examining footage from ten cameras covering an entire alleyway.
With a video analytics engine, the investigative work could be dramatically shortened. If the security agent knows details about the perpetrator being targeted, he or she can search the video, filtering based on the suspect’s attributes and known features, and eliminate from the video search all objects that don’t match the description. Instead of watching hours of footage, the investigator can narrow down the footage to the appearances of relevant objects similar to the suspect, and quickly identify the necessary video evidence for build- ing a case.
By accelerating the video review process, police forces can dedi- cate fewer officers to extracting video evidence, while collecting more of it by expanding the scope of the video investigation.
In cases without a known suspect or little direction guiding the video search, streamlining video review is even more critical. By de- tecting, extracting and classifying video objects to understand the context of a scene, AI-driven video analytics technology creates a structured information database out of the unstructured video data. Thus, investigators can quickly and comprehensively evaluate all vid- eo events and respond based on actionable intelligence from surveil- lance that otherwise would be underused.
Machine Learning for Driving Real-time Response
Security doesn’t always involve post-event investigation—much of the work behind ensuring public safety, revolves around respond- ing to events as they unfold. From uncovering a suspicious detail, to identifying a potential threat and deploying responders, dangerous incidents can be prevented, and security breach damage can be sig- nificantly curtailed by real-time response.
Through Machine Learning techniques that train video analytics to recognize patterns, Video Content Analytics solutions provide an invaluable tool for detecting anomalies and suspicious behavior. With AI-backed Video Content Analytics, law enforcement and security professionals can configure notifications to be alerted when unusual behavior may warrant their response. This is an essential capability for managing access control, preventing trespassing and monitoring loitering. When certain areas under surveillance are defined as sensi- tive, a call to action can be triggered any time an object enters or dwells in that area.
However, the security value can be extended even further for emergency response. At a hospital, for instance, alerts can be config- ured to notify security whenever ambulance access to the Emergency Room is blocked. The same technology can be leveraged by a munici- pality or local police force to ensure vehicles aren’t obstructing fire hydrants or emergency vehicle access in public places. The ability to proactively respond when a car is blocking the way of an emergency vehicle can be the difference between a patient receiving emergency treatment in time to save a life or the critical moments a fire fighting team requires to combat a blaze before it spreads.
Extending Human Detection with Computer Vision
While an attentive detective easily can notice abnormal behavior in a live video feed, there are many details captured by video surveillance that aren’t visible to the human eye. With Computer Vision technol-
ogy, these objects can be detected and indexed with the rest of the video data by the Video Content Analytics engine. Whereas a security officer in a control room monitoring a VMS may not be able to iden- tify a shadow or reflection of an object, a video analytics solution could be configured to varying levels of detection sensitivity.
On the surface, a video may not prove the presence of a suspect at a crime scene, but a video analytics engine, set to the highest degree of detection sensitivity, might identify a perpetrator’s reflection and extract it as a video object. While the criminal was outside the surveil- lance range, if his or her reflection appeared in the surveillance foot- age, it could be logged as video evidence to support a case.
This is not only true for detecting objects, but also for analyzing events and drawing connections and conclusions about recorded data and incidents. Machine Learning enables the collection and process- ing of data in ways human analysts cannot, presenting it for easy consumption and interpretation. With deeper data insights, security personnel can use otherwise unutilized but valuable information for constructive applications.
Video Data-driven Decision Making
Video Content Analytics renders raw data into structured data. When organized into dashboards and visualizations, the intelligence easily can be analyzed by security forces. With an additional intel- ligence layer, security personnel benefit from a high-level overview of all surveilled objects and sites but also from the technology’s ability to correlate between these data points and provide added insight.
For public safety officers, for example, this Smart City technology can be leveraged to help drive traffic optimization. The Video Con- tent Analytics engine can identify the number of vehicles traveling in every direction and understand high dwell time locations and du- rations. Knowing that eastbound traffic experiences increased dwell time at certain stoplights at specific times, the city can take action and optimize traffic flows using the video surveillance infrastructure al- ready in place for monitoring the roads. Traffic control officials might never have noticed the connection between all these data points, but, when in the context of a dashboard visualization, this intelligence can be leveraged to effect impactful change.
Video Content Analytics provide cities and local law enforcement with the tools to optimize resident lifestyle beyond public safety. It enables them to measure the efficacy of public infrastructure, trans- portation services, and urban landscape. Another example of this would be leveraging video data to make informed decisions and en- able intelligent planning of bike lanes, based on identifying the fre- quency and high concentrations of bikers on main roads.
Airport security could leverage the same data visualization and reporting capabilities for optimizing security screening processes. Tracking patterns over time and understanding how the security check points are navigated, the security and operations professionals can identify the causes and locations of bottlenecks, prevent them from forming and formulate contingency plans for overcrowding for expected and unexpected influxes of people. Quantifiable data en- ables organizations to plan based on trends and even A/B test solu- tions to overcome challenges and meet internal and communal needs.
Video has always been a key sensor for enabling safety and se- curity, but Video Content Analytics introduce a deeper dimension for harnessing the power of surveillance: operations optimization. By offering access to data and the tools to respond productively, proactively and predictively to situations, secu-
rity agencies can discover inefficiencies and their causes, streamline investigations and emergency response, and resolve diverse challenges as they develop.
Stephanie Weagle is the chief marketing officer at BriefCam.
36
1018 | SECURITY TODAY
VIDEO ANALYTICS