Our North American Editor at Large Rachel Levy-Sarfin continues her series on everyday applications of Machine Vision here with her article from our May issue around home security cameras. She covers topics from recognising faces to alerting us when a package arrives, and how machine vision has quietly become a fixture in our everyday lives. This powerful technology once the stuff of science fiction, is now in our pockets, our cars, and even watching over our homes
We use machine vision every day, even if we don’t realize it. It’s become embedded in our daily routines.
Last month, we kicked off a three-part series about the use of machine vision in our day-to-day lives. We discussed machine vision in cars, and this month, we’ll look at how machine vision is used in security cameras.
Security Cameras Over the Years
There’s some evidence to show that the first security camera was invented by a Russian scientist Leon Theremin (the same man who invented the musical instrument that bears his name) in 1927. However, the Kremlin quickly classified it.
Fifteen years later, during World War II, German engineer Walter Bruch designed what many people consider to be the first closed-circuit TV system. His invention allowed Nazi scientists and military personnel to safely observe the launch of V2 rockets. However, Bruch’s invention couldn’t record the images it screened.
Over the years, engineers refined CCTV further. In 1949, the earliest CCTV cameras became available to the public. Two years later, engineers created the videotape recorder. The earliest versions recorded live images reel-to-reel on magnetic tape, and they were expensive and bulky.
Governments and large corporations realized that CCTVs and videotape recorders had surveillance applications. British security services used CCTV at Queen Elizabeth II’s coronation in 1953. By 1960, London Transport installed CCTV cameras in a train station. New York City followed suit in Times Square.
Inventor Marie Van Brittan Brown saw the progress made with CCTV cameras and invented the integrated home security system. It included four peepholes, a sliding surveillance camera, a monitor, and two-way audio.
Eventually, other advances in technology made it possible for individuals to install home security cameras as part of wider security systems. As time passed, manufacturers made security cameras smaller and less noticeable. They also improved their effectiveness by adding features such as night vision.
By the late ‘80s, manufacturers began releasing digital cameras. These cameras no longer needed cassettes to record footage; they used hard drives and flash storage. Eventually, internet protocol cameras came onto the market. They could wirelessly transmit footage over a computer network.
Machine Vision and Home Security Cameras
One of the developments of the Digital Age has been machine vision. Machine vision allows cameras to detect specific objects, such as people. Sensors send alerts to the home security system to notify owners that someone is outside.
The first security cameras with facial recognition hit the North American market in 2012. Over the years, it became an industry standard. Object recognition has also become an industry standard, allowing cameras to detect when you receive a package on your front porch.
The Evolution of Machine Vision in Security Cameras
The integration of machine vision into security cameras, especially for consumer use, has been an evolution.
When this technology first came out, home security camera users would receive false notifications that someone was at the door. However, the technology has improved so that cameras can more reliably detect people and objects. The days of false notifications have passed.
In fact, today’s home security camera owners can filter notifications so that they only see what matters most to them. If they’re worried about animals in the back yard, they can set up their camera system to alert them when a squirrel tries to attack their bird feeder.
In addition, the image quality has also improved significantly. Images are crisp, so users don’t have to guess if it’s the delivery person at the front door or a potential burglar.
Moreover, all of these capabilities were previously available on much more expensive, premium systems. Advancements in cameras and machine vision technology mean that manufacturers can incorporate that functionality into even the smallest cameras. Renters can install peephole doorbells—they don’t need to invest in an expensive, comprehensive security system.
The Future of Machine Vision Technology in Home Security Cameras
While manufacturers have made enormous strides in improving image accuracy and notification filtering, they’re still working on enhancing security camera functionality.
Some manufacturers are working with advanced model training to teach machine vision systems to recognize movements typically associated with burglars and other criminals (such as loitering, skulking, covering their faces, etc.). This algorithm training helps cameras differentiate between bad guys and benign visitors, like the Uber Eats driver.
Additionally, some manufacturers offer digital forums to discuss camera footage with other users. For example, if it’s prom night, someone in the neighborhood can let everyone know that those groups of people wandering around in the dark are just teens coming home from their big night. As technology improves, it will most likely become easier to share updates with the neighborhood about what’s going on as new devices come onto the market.
And when it comes to devices, we can expect machine vision-enabled cameras to integrate with other networked devices in the future. Experts predict there will be an ecosystem of smart devices linked to one another, including security cameras, increasing convenience for users.
To increase convenience, industry observers predict that we’ll see more summaries of what cameras record. Users will be able to tell at a glance what the camera has picked up and if they should be concerned.
Additionally, we can expect greater use of voice commands in home security monitoring, including home cameras. Someday in the near future, you’ll be able to ask your home monitoring system if your teenage son brought in the garbage can. The camera will pan down the driveway to check for the presence of the can and show you the footage.
The Risks of Machine Vision in Home Security Cameras
As with any technology, the use of machine vision is not without its risks. At the time this article went to press, machine vision in home security cameras is still in development. While it’s been incorporated into home security cameras for over a decade, it hasn’t completely matured.
There’s still bias in the models used to train machine vision systems in cameras—some models are less likely to flag a video for police intervention when the residents are mostly white
There are also privacy concerns. The camera could capture someone’s image, and the manufacturer could potentially use that data as part of a model for AI training. The person being recorded never consented to that.
While we can enjoy the benefits of greater security, we must also be aware of the risks. We also can’t be complacent, and we shouldn’t be afraid to ask questions of manufacturers about bias in AI models and how their models are trained.
Machine Vision: An Integral Part of Our Lives
Every time our doorbell camera sends us an alert someone is at the door, we’re using machine vision. It’s become so embedded into our lives that we don’t even think about it, yet it works around the clock to keep us safe.
In the next and final part of the series, we’ll look at how machine vision integrates into augmented reality, and how we’re using it on a daily basis.
