.jpg)
Image CreditS: ra2studio/shutterstock.com
With new surveillance systems and smart sensors heading up the Internet of Things revolution, the extent to which you can monitor your home remotely is ever expanding.
The most widely used form of sensor in home security is the classic burglar alarm, triggered when somebody enters the house. A form of the burglar alarm was pioneered in 1853 by Augustus Russell Pope, who formed an electro-magnetic alarm system by simply connecting doors and windows as an electric circuit – when the circuit was closed, the current caused vibrations in a magnet, which itself was transmitted to a hammer. This hammer then struck a bell, informing inhabitants that the circuit had been closed.
Motion detector technology has evolved somewhat from 1853, with modern, everyday systems available from most hardware shops for a fairly inexpensive price.
The Lastest Technology
However, newer, more sophisticated technologies may be beginning to take the place of our standard motion-detector burglar alarms. ‘Strips’, a crowdfunded project, is an invisible magnetic sensor that can detect whether a window or door is open or closed. It can transmit this data to the user's mobile phone through the Z-Wave gateway, allowing users to remotely check that their home is secure.
Sensative, the company behind Strips, hope to introduce a vibration sensor into future models of the sensor to allow the user to be notified if the window it is attached to is broken.
Taking things a step further, companies such as Nest, Simplicam, and D-Link, have developed ‘home surveillance cameras’, with users connecting to the devices via their smartphones.
The software can tell users when an unrecognized person is in their home, show pet owners what their pet is doing, or enable parents to keep an eye on their children – some similar cameras, such as those produced by Withings and Medisana, market themselves as ‘Smart baby monitors’, giving parents the opportunity to watch, talk to, and zoom in on their baby’s face.
Who is in Your Home?
Facial detection software has been a component of basic digital cameras for a number of years. Most modern cameras can detect whether an object in the viewfinder is a face, and edit its settings to take this into account.
The Viola-Jones framework effectively identifies faces with a high rate of accuracy, with false negative rates of less than 1 percent and a false positive rate of under 40 percent. When an image is defined as a set of intensity values (from 0 to 255), differential blocks can be identified, reducing the complexity of the face into lighter and darker ‘blocks’.
Systems can then be trained, using databases of both faces and non-faces, to recognise those ‘features’ that are typical of human faces. Using an AI technique called AdaBoost, a cascade with 32 stages was produced, with 4,297 informative features out of the possible total of 45,000. This method simplified the process of recognizing a face, with modern cameras able to do this much faster than previously, and without needing much power to do so.
Facial recognition, a beta feature currently offered by Simplicam, is the narrowing down of digital faces to potentially match one of those in a database. This is a much more complex field of study. The first semi-automated systems for recognising faces were developed in the 1960s, but these techniques required manual input from administrators, such as the location of particular facial features.
Video Credits: Nest/Youtube
Along with facial recognition, Simplicam is also equipped to differentiate between animals and people through heat sensors, so that animals outside or pets inside the house are not mistaken for human intruders.
History and Developments in Facial Recognition
In 1988, a technique called Principle Component Analysis (PCA) was developed. PCA essentially reduces the dimension of data through data compression, with faces represented as a feature vector of a set of basic features, or ‘Eigenfaces’. However, the effectiveness of this approach depends on environmental factors, such as luminance and the angle of the face.
Linear Discriminant Analysis (LDA) is a statistical approach that focuses on minimizing variance across users and maximising the differences within user variance, classifying faces based on known training samples.
Another research direction is Elastic Bunch Graph Matching, which uses ‘bunch graph’ approaches to compare images based on elastic grids that produce values to represent points on the face, such as the eyes or nose. This approach sidesteps many of the difficulties brought up by variety in facial expression, position and pose by effectively extracting and comparing values, using a process that may be used in the visual cortex functioning of some mammals.
3D facial recognition can help to identify faces that are in profile, rather than facing forwards. Commercially-speaking, companies such as Google, Toshiba and Samsung, as well as research bodies such as Cambridge University and Carnegie Mellon University, have all been involved in The Face Recognition Vendor Test (FRVT), which tests algorithms in order to find effective means of distinguishing the identity of a face.
The 2013 FRVT report evaluated the effectiveness of techniques including age estimation, gender estimation, finding differences between twins, and the verification of visa images. Facial recognition software has obvious functions relating to security.
As well as home security cameras like Simplicam or the in-development Buddyguard’s Flare device, the technology could also be used for access control – for example, to offices or computer systems, and for identity verification, surveillance, and image database investigations.
Functions outside of home surveillance through images include home security hardware that monitors water levels in bathrooms or in kitchens to notify homeowners of potential leaks, as well as keeping track of smoke and carbon dioxide levels in the home.
Although elements of the technology are still being developed, it is clear that the hardware related to home security is rapidly advancing. Developments in sensor technology have been aided by advances in our utilization of wireless networks, more efficient battery power, and the continuing ‘Internet of Things’.
This article was updated on the 1st August, 2019.
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.