It’s no surprise that security cameras have advanced with the progress in technology. The first ever closed-circuit television (CCTV) system was developed during World War II in Nazi Germany, so scientists and engineers could watch test launches of the V-2 rocket. It took the introduction of videocassette recording (VCR) in the 1970s, however, before video surveillance could become commonplace.
Of course, the grainy images of those VCR-based systems have been replaced by sharp, high-resolution recordings. What’s more, while those older cameras required networks of coaxial cables to be installed, WiFi security cameras can send their recordings wirelessly to a central server. In addition, a wireless security camera system can be monitored from anywhere in the world with an internet connection – given sufficient security precautions, of course.
With a modern security camera, wireless technology is commonplace and their features are developing rapidly, so it’s only natural to wonder how they could possibly be improved further. Automation through artificial intelligence is one area where the security cameras of the future could surpass today’s technology.
For example, you could install a home camera and record everything that happens on your property, but you would probably only check the footage if an incident occurred. Even then, you may need to process hours of footage before you find the right scene. Modern home cameras often have a motion detection feature. This triggers an alert, usually by email or a companion smartphone app, when something in the camera’s view changes (according to a set sensitivity). While this clearly helps reduce the volume of data that needs to be processed, it can also capture innocuous activities like postal workers and paperboys going about their duties.
To address this, home cameras of the future may come equipped with deep neural network artificial intelligence. Rather than relying purely on motion detection, these neural networks will be able to identify suspicious behaviour and send a real-time alert. The idea is that you may then be able to stop crimes before they happen, or at least increase the chances of the police being able to apprehend the perpetrators.
To give an illustrative example, imagine you are dining out one night when you receive an alert on your phone. In the accompanying video, you see someone trying to break in through your back door. You can now quickly report this to the police, who then have a chance to catch the would-be burglar in the act. In addition, if you have home automation, you could activate something that may startle the intruder and deter him from continuing, such as turning the lights on and playing loud music.
The problem of processing data is a particular problem for the security services in London, and the way they address it may give us some hints about how our home wireless IP cameras may evolve in future. For example, following the riots of August 2011, the Metropolitan Police had to manually analyse over 200,000 hours of recorded footage. Their months of effort were rewarded, however, by the identification of around 5,000 offenders. This shows the value of CCTV in the fight against crime, but clearly the police would prefer that it did not need so many officers, who also need to be trained in CCTV analysis, to go through the footage.
The government realised that the outdated tools the police were using needed to be replaced, so it announced the Video Analytics for Law Enforcement (VALE) programme in 2015 to develop new tools for the security services to use when analysing CCTV.
WIRED reported how it visited a company developing new video analysis tools for the police forces of the UK. Given the sheer number of CCTV cameras in the United Kingdom (estimated at as many as 5.9 million), the approach was taken to process the footage from “dumb” CCTV cameras rather than propose installing new cameras with advanced analysis abilities. The first step is therefore to convert all the available CCTV footage into MPEG4 format. While this may seem mundane, the police previously had to cope with over a thousand video formats from different camera systems.
The first tool that was demonstrated involved some CCTV footage from the 2011 riots in London. In the scene, the recording is focused on a door that someone is expected to come out of. Rather than go through hours of recording before finding the correct footage, the software analyses the video and classifies motion scenes into high or low probability depending on the subject’s proximity to the camera. This means an officer can then just analyse the scenes where motion was detected.
The second tool involved facial recognition in crowds of people. The reporter was shown footage of people exiting a London tube station. The security cameras system then analysed the recording for faces and displayed the different faces on the side of the screen, grouping different instances of the same person together. When an operator then clicks on a face, he or she can watch what that person was doing at the time.
While catching a crime on camera is immensely useful for identifying and prosecuting the offender, you also want to know what that person was doing before and after. For example, did that person have any accomplices, or did he or she later climb into a vehicle with a visible registration plate? The third tool helps with this by allowing the operator to highlight a person on a piece of footage and have the system find other recordings of similar people. While facial recognition is used to rate the probability of a match, a human operator still needs to make the final determination in each case.
It would not be surprising if similar technologies made their way into home cameras. For example, facial recognition could allow a camera to learn the faces of “trusted” visitors (e.g. you and your family) and prevent unnecessary alerts from being generated.
It is, of course, hard to predict how WiFi security cameras will evolve in the distant future. One day, we may even see artificially intelligent drones flying themselves around our properties and detecting crimes before they happen. Maybe they will even be networked with the local police, informing them directly of crimes in progress or suspicious activities, and allowing the authorities to respond instantly.
Added by @time2hq on 2nd Aug 2018 in Security Cameras
October is National Home Security Month, the aim of which… Read More
It is surprising how many features can be found in… Read More