Video AI helps to produce business intelligence
Source: John Liu
The use of smart cameras
with embedded artificial intelligence (AI) is gaining traction in the security industry. A number of companies at ISC West showcased products utilizing the technology.
allows for analytics to be performed on a camera. The technology is being deployed in a number of environments, including parking lots, retail stores and airports. Embedded AI in cameras is trained via machine learning
. A large cache of images is fed into a system, from which AI is able to learn and detect poses and objects. Consequently, a device is able to recognize the distinctive features of people, as well as objects such as cars, boats, guns, and bags, depending on the requirements. Normally it takes about 30 days of training for a camera to learn the shape of a new object. The consensus among those interviewed at ISC West was that devices were now able to perform accurate identification more than 90 percent of the time.
These types of smart cameras
are often deployed in the transportation vertical. An example is mounting multiple cameras at intersections to check on traffic f low. The cameras are able identify the type, speed and direction of passing vehicles, as well as times of peak congestion. At ISC West, Bosch
also demonstrated how its video analytics can be used in parking lots. Bosch's camera can monitor vehicles moving in and out of the parking lot, and calculate the total number of cars in real time. The camera can also identify the number of vehicles in handicapped, electric car or high-end car sections. Intelligent Security Systems showcased its AI-equipped camera, which can detect and time cars stopped in the drop-off lane at airports.
has gained widespread adoption not just for security purposes, but also as a means of gaining business intelligence. Bosch's camera with facial recognition offers a multitude of features for retail stores, such as people counting. Stores can use this to determine if there are enough staff, where most of the foot traffic comes from, and which doors or entrances should be monitored more closely.
Blocking out faces is a common feature adopted by many to meet EU General Data Protection Regulation (GDPR) privacy rules. Honeywell's video analytics solution is one example. Its internal algorithm can tell whether the person identified on the camera is known or unknown by the system. If the person is unknown, a blue box will block out his or her face on the screen. On the other hand, if the person has been previously identified and enrolled in the system, their face will be shown.
Vintra, another video analytics provider, also showcased its facial recognition solutions at ISC West. Vintra's device identifies faces with different confidence levels shown on the side. This information is helpful for police or internal investigations, both live and post event.