Page 58 - Security Today, August 2017

SMART CITIES
pixels, or a portion of the pixels, to identify suspicious objects and people and an alert would then be sent so security teams could investigate or respond to the threat. The chal- lenge with this type of technology is they could be easily tricked into falsely identifying an animal or object as a suspicious person.
For example, an alert that someone had entered a highly-restricted area monitored by an outdoor camera that could easily be triggered by a squirrel or raccoon that came into the camera’s field of view. The software would simply detect a change in pixel move- ment or colors and then trigger an alert— there was no intelligence behind it. All the algorithm knew was that something new was in the scene, but the ability to identify whether the intrusion was human, assess its threat level and determine if it warranted of action was missing.
Early video analytics technologies also created many false positives due to adverse weather conditions, like snow, wind and oth- er daily environmental evets, which also trig- gered an inordinate amount of false positive alerts. People tend to start ignoring the alerts that cry wolf, essentially making the technol- ogy useless. These types of false positives would waste valuable time and resources, or in many cases, caused security personnel to ignore alerts. This ultimately slowed market adoption of video analytics. Video analyt- ics that are useful need to provide critical situational awareness in all environmental conditions—even adverse weather—and ig- nore irrelevant intruders like squirrels or the neighborhood cat.
Today, computer vision, machine learn- ing and artificial intelligence are dramati- cally lowering the rate of false positives and advancing the usefulness of video analytics. Rather than just seeing moving pixels, ad- vanced video analytics can create 4-D recon- struction of two-dimensional video images (3-D plus time). Now, not only are objects able to be detected using particles, perspec- tive, velocity, path deviation and travel dis- tance, they can be accurately identified and in some cases classified through artificial intel- ligence. Security personnel can leverage these advancements to better understand if a per- son or suspicious object has been detected.
When a squirrel runs through a video feed, advanced video analytics capture mul- tiple angles of the animal. On the back-end, it takes those two-dimensional images and reconstructs them to build a 4D version of the squirrel. The object is now identifiable against a database of known images—each with an assigned level of potential danger. The video system can not only see an ob- ject, it understands what that object means in the context of the area being monitored: a squirrel is only a squirrel, not a trespasser that might be armed and dangerous.
Unfortunately, not all objects are as inno- cent as a passing squirrel.
How many times have you been at the airport and heard the loud speaker come on reminding passengers to keep their luggage close and report any unattended or suspi- cious items? That’s because airports are a constant point of public safety concern. With thousands of people shuffling between gates and destinations—luggage and bags in hand—any passenger could be a threat or could have just forgotten their luggage in the confusion of travel.
So, how do you keep people safe when there is so much going on? Arm them with enhanced awareness, security intelligence, and the ability to better allocate resources.
Say a bag has been left unattended. With today’s advanced video analytics, the object would be recognized as a bag as soon as it enters into the purview of a video camera— then the clock starts ticking. Depending on the procedures in place at the airport, staff would be alerted once the threshold of desig- nated time had passed that there was an ob- ject that has not moved or interacted with a passenger for longer than the allocated time. The video system could then send all first re- sponders, in this case airport security, a real- time alert to their mobile devices, notifying them a bag has been abandoned, its location and how long it’s been there. Now everyone is on alert and the security team can immedi- ately determine who should investigate based on geo-location. A plan of action can be formed in real-time, based on real insights.
If the situation goes beyond a passenger taking too long in the restroom, advanced video analytics can immediately arm inves- tigators with the situational awareness they need to begin an investigation, while con- tinuing to keep everyone safe. Through fa- cial recognition, the software can also iden- tify all the people who have come in contact
with the bag—going so far as to backtrack all their steps from the point of abandoning the bag. Leveraging the airport’s parking lot cameras, a suspect’s car and license plate can then be identified. With a visual of a suspect and vehicle information in place, law officials can begin to investigate any criminal connec- tions and evaluate if the bag might pose a bomb-level-threat worthy of evacuation.
While this might be an extreme scenario, the reality is that when a major event or ev- eryday situations occur, we must prepare to stop them—and advancements in computer vision, artificial intelligence and analytics have been helping law enforcement and op- erations teams tackle these challenges more and more. For example, video analytics can also detect when lines become too long and alert staff to open another ticket counter, security gate, or passport stand in order to help people get from the door to their gate as quickly and conveniently as possible.
Smart Cities Fuel These Superpowers
For complex organizations like governments and cities that produce massive amounts of data, video, while valuable, is greatly en- hanced when integrated within a broader smart city ecosystem. By connecting disparate video, data and public safety systems on a sin- gle platform, public safety officials can receive real-time alerts backed by the most up to date insights, map and predict crime to deploy re- sources and strategize the best approach to a crime that’s in progress or how to prevent the next one—maximizing the level of safety.
Take the airport scenario. While video analytics may help set up police officers for success in identifying a suspect, integrating the system with a broader smart city plat- form will expedite results. Rather than hav- ing to wait for a suspect’s image to be located in a law enforcement database and try to
monicaodo/Shutterstock.com
42
0817 | SECURITY TODAY