Video surveillance has advanced to the point where cameras are beginning to double as smart motion/intrusion detection sensors. Many can alert users to potential security problems at the protected premises and enable them to view activity in real-time from connected devices. Detection is automated and accurate, so there’s less dependency on manpower, and central monitoring operators can quickly focus on critical events and the most serious security threats.
Steve Burdet, manager of solutions management at Axis Communications, Chelmsford, Mass., discussed what’s driving this evolution.
With all the innovation in video technology, will cameras become motion/intrusion detection sensors in an all-in-one solution?
Video technology continues to improve in low and dynamic lighting environments and in terms of higher resolution. All of this improves its function as a detector. That said, as with any technology, it has its limitations. Namely, if it cannot see, it cannot detect.
If the environment allows the camera to see at all times required, there is a strong case that these devices can be extremely effective motion/intrusion detection sensors. With the advancements in A.I. [artificial intelligence] analytics, cameras have become much more reliable sensors to capture only what you need and with a high degree of accuracy. This reduces nuisance alarms, increases the forensic value of video and allows users to confidently respond to events they care about.
If the environment does not afford this (i.e., nighttime does not provide enough light to ensure detecting at required distances or accuracy; and fog, snow, sand, smoke, rain, etc., are other variables that can reduce a camera’s visibility), then additional devices would be necessary. For example, a fusion camera that blends radar and video technologies can leverage radar to see, regardless of light. It can provide that information back to the camera to record or provide an overlay of the detected object on the video (even if the camera cannot detect the object still). It could also be used to activate additional lighting to allow the video to capture the scene more clearly.
What makes this transition possible?
Advancements in three areas—physical camera design, imaging and analytics (using A.I.)—are the main contributing factors. Based on these elements, automation and intelligent systems simply respond as directed.
First and foremost is the physical camera design. Good data in equals good data out. Surveillance cameras must capture video in a wide variety of environments, including extreme weather or temperatures, vandalism and vibrations. If there is condensation on the lens or vibrations preventing the camera from working, then you acquire bad data that’s unreliable. Hardware sets the foundation that all other elements are built upon.
When it comes to imaging, improvements to the image tunnel are key. The image tunnel refers to the image processing from sensor to digitization. This is where color representation, noise reduction algorithms and technologies like wide dynamic range are applied. These elements allow cameras to blend the advancements in image sensors with advancements in edge computing to make the raw data into the best possible and most usable image. With additional computation, we can run more processes in parallel or more advanced versions of the processes.
In addition, when talking about the transition toward an all-in-one solution for detection, analytics are key. When we have well-designed hardware providing good data that the image tunnel is processing, an analytic can be used to analyze data for specific things such as classification (people, cars, etc.), attributes (color, style, etc.) or event (direction, size, time duration, etc.)—even a combination of these things.
And with A.I., analytics have become much more accurate. Prior to using A.I., shadows, light changes, insects and other factors would often trigger motion, making sensors less reliable. This lessens confidence and the ability to respond to an event properly. By classifying objects, you can filter out the things you do not care about. These advancements and capabilities immediately changed the conversation about using analytics as a detector and even for broader purposes. The confidence in the data that A.I. has created has been a game-changer.
Will stand-alone detectors move to more specialized applications if this transformation continues?
I believe the market suggests that there is still a need for specialized sensors tailored to specific needs or to be deployed in environments where cameras cannot be used. Along with the use cases like environmental sensing, vape detection and the growth toward loT and smarter, interconnected systems, there is increasing interest in sensors of all types which can create useful data for business intelligence, operational efficiencies and more.
steve burdet
