Facial recognition camera – the next step in home security
Facial recognition has become a common feature of the modern world. It could be to unlock your smartphone, a Snapchat filter, or for airport security. Facial recognition camera is finding wider use and is here to stay. Reports claim the facial recognition market will grow to $9.6B by 2022. With a compound annual growth rate of over 20% between 2016 and 2022. So how is this tech finding use in our sweet home?
Facial recognition cameras for homes are mainly used for security. But they can also be used to automate processes and for family monitoring. Companies such as Google Nest, Hikvision, Netatmo, and SimCam already offer smart home devices with face recognition.
How does facial recognition work?
Human beings are very good at remembering faces. We’ve all had that feeling of knowing you’ve met someone but not knowing their name. You know their face and you may have met them at work or a party, who knows. Their name and how you met is gone, but you still know their face. Inside your head, you’ve stored the mix of features that are unique to them. Turning a face’s mix of features into biometric data is precisely how facial recognition works.
1. Face Detection
This could be from an image or video of a single person or a crowd. It works best when the person is looking directly into the camera. Upgrades are being made with detection possible for faces turned slightly as long as key features remain in view of the camera.
2. Face Analysis
The software reads the face finding landmarks or nodal points. A human face has roughly 80 nodal points. Mapping these nodal points produces key data such as:
- distance between the eyes
- width of the nose
- shape of the cheekbones
- length of the jawline
- distance from forehead to chin
Combining this data builds a unique digital faceprint, similar to a fingerprint.
3. Face Matching
Finally, a database search of known faces finds if there is a face that matches. Accurate facial recognition needs a large database of known faces. A report in May 2018 claims the FBI has access to 412 million facial images for recognition searches. These could be drivers license images or photos posted to social media. User-created databases of images on social media provide a large source for facial recognition. These images often have each person within the photo tagged.
Facial recognition cameras for home use allow the creation of a database of known faces. Devices such as the Nest Cam IQ can find unknown faces in a live video then alert the owner. Known people can have a profile created on the spot. For other devices like the SimCam 1S, to create a profile a clear photo has to be entered.
What are the uses of facial recognition security cameras?
The home use market for facial recognition is growing. This could be for both outside use and indoor use. Smart home facial recognition provides users with improved security and convenience. Below are some key uses for smart home facial recognition.
Spot strangers on your property
Home surveillance with facial recognition allows a security team to spot and respond to suspicious, threatening, or banned people. In the event a crime occurs, these security systems provide crucial evidence to police officers. This can lead to faster justice and a higher chance of recovering stolen goods.
They require the user to set up a database of known people. Upon the detection of an unknown face, alerts are sent to the owner, and a video of the event is stored. This is mainly for use against would-be criminals. Other uses could be to tell the owner when a family member brings a stranger home. It could even be set up to produce alerts showing when a delivery driver drops off a package.
Outside of security, monitoring the people living at a house is a key feature. This includes telling parents when their kids arrive home, or alerts when they enter restricted areas that could lead to injury. If you have dinner slow cooking and want to make sure your child doesn’t burn themselves or if you have a home workshop that is not fit for children.
Facial recognition cameras can also monitor older relatives. Combining them with advanced AI video monitoring software could even send alerts if they were to fall. Another use is tracking other visitors to the house such as nannies, caregivers, cleaners, and repairmen.
Home automation via facial recognition
Imagine arriving home from a long day at work. A camera detects your face and the smart lock opens the door. The lights turn on and your favorite playlist starts playing. Whatever your routine is it can be set to trigger based on facial recognition. Each family member could have their own routine set up.
With advances in AI technology and smart home control systems this level of automation is becoming a reality. Security cameras such as the SimCam 1S can already be combined with Sonos to play one’s favorite song upon detecting known faces.
The problems with facial recognition
Although facial recognition is finding wider use there are still problems that need improvement, these include:
The key pushback comes from people with privacy concerns. Many people believe the storage of biometric data and global databases of facial recognition images are an invasion of privacy. Also, this data could be sold to third parties or become a target for cybercriminals. Recently facial recognition software firm Clearview AI reported a data breach. Cybercriminals gained access to the company’s data including the customer list and the number of searches each customer had made.
These privacy issues are focused on mass surveillance systems. However, there are potential hurdles for home use facial recognition systems. A customer setting up a facial recognition camera system is consenting for it to store images of themselves. However, problems can occur with data obtained from other people. In America and many other countries, it is legal for a homeowner to record anyone on their property. Though It may cause issues with any contractors working at the property.
To remove some of these concerns, SimCam uses local storage and local processing for face recognition. Sharing personal data with third parties is impossible on this type of device.
Facial recognition inaccuracies
Today, facial recognition algorithms can have an accuracy of 99.97%. But that’s in ideal conditions. The accuracy can be affected by certain factors. This could be lighting, the distance from the camera, and the angle of the face with relation to the camera.
AI vision usually can’t work properly in the night, which is the same for human vision. But this can be improved with an enhanced night vision infrared LED. Moving closer to the camera leads to clearer facial images and greater accuracy. According to Digital Trends, a military-level facial recognition device can identify people from 1km away. For a consumer-level security camera, the best distance is around 5 meters.
Facial features greatly change with the face angle, compared to the registered data. Computer scientists have addressed this issue with data augmentation. For example, SimCam’s AI software can automatically map out facial features at different angles. The camera can recognize the owner’s side-view faces by registering only one front-view photo.
Facial recognition bias
Studies have shown that current facial recognition systems have greater accuracy when identifying Caucasian male faces. Systems have been found to falsely identify Black and Asian faces at higher rates. There are also more problems identifying female faces.
The National Institute of Standards and Technology (NIST) tested 189 facial recognition systems from 99 developers. This included software from Microsoft, Cognitec, and Megvii. Amazon, Apple, Facebook, and Google did not submit their software for the federal study. It found false identifications were 10 to 100 times higher for darker-skinned faces.
In fact, AI has no bias at all, neither does the engineers who develop the software. It is merely because even the best AI algorithm has difficulty in reading facial features on darker skin or a woman with heavy facial cosmetics.
Fooling facial recognition cameras
Another issue is would-be criminals fooling facial recognition security systems. Studies have shown high-quality photos of the owner can unlock many devices using facial recognition identification. Dutch consumer protection group Consumentenbond tested 110 phones and found 42 failings. Most facial recognition technology is based on 2D images. Smart home cameras like Nest Cam can pick up “unknown faces” on TV screens and trigger many false alerts.
Perhaps a greater problem is the use of 3D printed masks. Research has shown that 3D printed masks showing someone else’s face can fool facial recognition systems. AI firm Kneron used lifelike masks to fool payment systems in China and a passport-control gate in Amsterdam. Currently, there is no fix to this problem even if a camera uses a 3D sensor.
Facial recognition technology for home use is a growing market. Used primarily for security, other uses include family monitoring and smart home automation. Many more home facial recognition cameras are under development. Although there are problems regarding privacy and inaccuracy the design of future systems is aiming to reduce these.
Home is where we need to feel the safest and facial recognition home security systems are the next step to providing that.