From healthcare and sports to eCommerce and banking, AI solutions have already achieved new milestones and brought considerable benefits. Yet, AI potential is still to be unleashed in a wealth of industries and business domains.
To wit, Accenture has recently revealed that enhancing CCTVs with AI and analytics is set to redefine the public safety ecosystem, reducing crimes in the USA by around 81.7 million by 2040. But what practical steps should be taken to turn AI into a vital cog in the public safety machine?
Tapping into face recognition
AI algorithms could become a vigilant partner in fighting crime. Powered with advanced face identification, verification, and tracking, AI-enabled systems can collect data from hundreds of thousands of live camera feeds and conduct a thorough analysis of faces and their attributes like gender, age group, hairstyles, hats, etc.
Such an analysis enables detailed face images comparison against criminal databases in a fraction of a second, helping the police slash investigation time and remove threats off the streets.
However, to bring tangible results, AI algorithms should be trained to operate under challenging conditions like varying facial expressions, shaky camera footage and motion blur, poor lighting, and more. Also, particular attention has to be paid to false positives and bias as well as face spoofing and morphing.
Leveraging object detection and tracking
High-precision detection and tracking of static and moving objects as well as comprehensive object profile creation are another step into the future of safe cities.
AI-fueled urban surveillance allows identifying suspicious objects including weapons, abandoned bags and their owners, masks, and intruders as well as hazards like smoke, fire, vibration, explosion, gunshot, etc.
To effectively perform object detection in complex scenarios — like occlusions, object deformation, and crowded scenes — and send timely warning notice, the public safety system should effectively use video timeline dependencies.
Making the most of vehicle identification
Taking object detection to the next level, AI-powered safety systems can be trained to precisely recognize various vehicle types like cars, scooters, bicycles, etc. and monitor their movements in real time despite poor lightening, high speed, and adverse weather conditions.
Coupled with a comprehensive traffic analysis and abnormal driving behavior detection, such monitoring can be instrumental in identifying drink-driving, speeding, illegal car parking, and other violations.
High-accuracy optical character recognition (OCR) is another reason why AI piques the curiosity of public safety experts. This sophisticated technology enables accurate recognition of car license plates, which significantly increases the speed of finding stolen vehicles.
Applying the power of behavior analysis
Underpinned by an advanced behavior analysis, AI is revamping public safety. Based on the results of object pose and activity estimation in public places, its cutting-edge mechanisms perform dominant and rare behavior mapping, which is key to spotting any deviation from the normal behavior.
Previously trained on large volumes of CCTV data, AI-enabled solutions can automatically detect any suspicious activity, for example, a person repeatedly carrying suitcases, a group continually gathering at night, or people wandering around the car park and trying to open multiple doors. Other examples include pushing, hitting, kicking, loitering, intrusion, turnstile cheating, and object abandonment.
These and many other suspicious activities are triggers for activating an alarm system — to prevent and address violence, robbery, suicide, shooting, and more.
It’s not all that simple
Equipping public facilities with CCTVs for real-time monitoring of suspicious objects and activities is costly, but it certainly gives more bang for the buck.
However, governmental bodies and developers of such AI-enabled public safety systems should consider possible fears of privacy invasion this technology might raise. It’s no out-of-the-blue idea given the situation in China, where CCTVs and data analytics are mostly used to rate citizens on trustworthiness rather than ensure public safety.
Author | Emily Forbes
An Entrepreneur, Mother & A passionate tech writer in the technology industry!