Secure the Environment

Secure the Environment

AI-enabled cameras are changing how campuses think about security

Protecting campuses, including schools, hospitals, business parks and distributed facilities, means managing an incredible number and variety of security sensors. This number gets larger every day. According to a recent Video Surveillance Report from IHS Markit, 85 million video surveillance cameras will be installed in North America by 2021.

With so many cameras, it is simply unrealistic to expect security staff to be able to monitor and manually search through the vast quantity of data being collected. Today’s campuses need solutions for managing their video data. They need AI, or Artificial Intelligence, to help them remain vigilant and identify when anything is out of the ordinary.

In addition to allowing campuses to continuously watch all their video streams, AI can also help operators focus their attention on what is important. By feeding data into systems that are able to sort, organize and categorize massive amounts of information in a way that operators cannot, AI can significantly reduce false alarms and enhance post-event investigations.

Deep Learning has Changed how we Use Security Cameras
Today’s AI, specifically deep learning, is changing how we think about and deploy security cameras. Deep learning allows us to extract more complex features from raw input. In image processing, for example, lower layers identify edges while deeper layers identify colors, letters or faces. Using supervised deep learning, we can now train cameras to recognize whatever we need them to see.

Deep learning algorithms do not see video the same way humans do. Rather, algorithms look for familiar shapes and patterns that they have been trained to recognize. For example, to assist with people counting, some AI cameras have already been trained to identify human faces and torsos within images. This frees security staff from having to count, and allows them to instead, spend their time on tasks only people can do.

It is important to remember that optimizing deep learning for security cameras is a sophisticated process. An algorithm that has been properly trained by exposing it to hundreds of thousands of images of objects with unique characteristics can be uploaded into an AI-enabled camera’s library to make comparisons and find matches. However, it does not have the capacity to “learn new things” by themselves.

This means that, if you want to train an AI-based camera to do something different or expand what it can do, you simply have to deploy new firmware that includes a newly trained algorithm. The benefit is that this allows manufacturers to take a lighter approach to AI hardware resources on their cameras while still providing campuses with the ability to deploy highly accurate, comprehensive security solutions to the edge.

Improving Security with AI-enabled Cameras
Campus security systems are collecting vast amounts of data. But, without the right tools, that data can overwhelm operators or be too much to manage. For security applications, there are two main areas where AI-enabled cameras can significantly improve operations: threat identification and forensic searching.

When it comes to protecting assets and people, real-time alerts generated by VMS can provide a good line of defense. Using motion detection, the system is able to identify potential threats in a camera’s field of view and then notify operators.

The problem is that this technology is not very accurate. It frequently generates false alarms by identifying anything from trees blowing in the wind to animals wandering through an open space as possible events. Because false positive alerts take up valuable time and resources, many campuses have decided to make minimal use of motion-based threat detection as part of their overall security strategy.

AI-enabled analytics cameras eliminate false alarms by accurately determining incidents that require further investigation by operators. Because they can be trained to detect and identify specific objects types, like people or cars, these cameras are able to recognize that a tree is not a person. This means that AI-enabled cameras eliminate the problem of false alarms because they ignore objects like wind, rain, shadows or even an errant plastic bag.

When it comes to post-event forensic searches, AI-enabled cameras also play an important role. These cameras collect metadata, including descriptive characteristics of objects like the color of a person’s shirt or pants, or their approximate age and gender. Using this metadata, the right VMS can quickly search through video to find a particular object or person. A search that might have taken security staff hours or days to complete, now takes only seconds when the analytics uses the embedded metadata provided by AI cameras.

How AI-enabled Cameras can Help Mitigate the Spread of COVID-19
Today, campuses are also using AI-based analytics and metadata to help prevent the spread of COVID-19. This technology is proving useful for monitoring social distancing and mask-compliance.

The ability to maintain social distancing in enclosed spaces is vital for mitigating the spread of the virus. But it is not always easy to achieve. Knowing how many people are in an environment with multiple exits and entries can be a challenge. Fortunately, AI-enabled cameras can support our efforts to keep campuses safe.

Using deep learning, Hanwha AI-enabled cameras can be trained to accurately count objects, including people, and determine their direction of travel. By counting people as they enter and exit spaces, these cameras, together with a compatible VMS, can allow a campus to display the occupancy of a room or building in real-time. When the maximum occupancy is reached, a “wait” message can be displayed on an overhead monitor to advise people from entering the environment. Similarly, when the number of people in the space goes back down to below the threshold, the system can switch the display to a “welcome” message.

Without AI, someone would have to manually count people entering and exiting each room. Using AI-enabled cameras protects security staff from possible exposure to the virus, and allows them to focus on other security tasks. This also helps people on campuses comply with social distancing policies.

Using AI to Monitor Masking Compliance
Mask-wearing compliance, especially indoors, can be vital in schools, business parks and hospitals. But security staff cannot possibly track every person throughout a campus environment to ensure that they are wearing a mask and have it on properly. Here again, AI can be trained to do just that.

AI-enabled cameras could already detect faces and torsos. So, it was relatively easy to modify the training to detect if someone was wearing a mask and if they were wearing it correctly. AI trainers simply had to tweak the existing algorithm that identifies faces and torsos to include data for facial coverings. The speed at which this was able to happen when it was needed is a testament to the agility of AI and why it is such an important technology for our industry.

Cameras like those in the Wisenet X and P series from Hanwha Techwin can now detect whether people are wearing masks, and the analytics can determine whether they are being worn properly. If the camera detects that someone entering an area is not wearing their mask properly, it can immediately broadcast an audio message directing persons to respect campus mask-wearing guidelines.

The AI can also trigger an event alarm in the access control and VMS to inform security staff when policies are being ignored. This provides them with a clear view into what is happening and allows them to restrict an individual’s access to certain areas if they are not following mask-wearing protocols.

Advances in physical security solutions means that we are collecting more data than at any other time in history. As this amount increases, we run the risk of being overwhelmed to the point of making it useless. Fortunately, AI-enabled cameras are changing the way that campuses can use their data to protect their environments. By identifying and categorizing information faster than any human ever could, AI technology helps security staff reduce risks and focus on tasks that matter.

This article originally appeared in the March / April 2021 issue of Campus Security & Life Safety.

Digital Edition