In 2018 we realise that Facebook is now doing facial recognition even on photos you're untagged in.
The myth of the surveillance society is nothing more than a myth. Just because there are security cameras in every street corner, it doesn't mean that it's infallible or that CCTV is delivering a safer society. When it comes to real-life situations, camera-based visual surveillance is not really accurate nor practical because you ultimately you need a human to watch the footage as they cannot rely on technology alone. For instance, during the 2011 London riots, facial recognition software contributed to just one arrest out of the 4,962 that took place. That is why visual surveillance still relies on people watching hours of footage - which is time consuming and unsustainable.
Software advances made in DNA sequence analysis could be a game changer in the field of video analysis software. These software tools and techniques, which treat video as a scene that evolves in the same way DNA does, could revolutionise automated visual surveillance.
While CCTV cameras create endless and complex video footage to analyse, automate video surveillance remains limited to tasks in relatively controlled environments. Although it is easy to detect a trespasser into somebody's property can be complete quite accurately, analysing footage of groups of people or identifying someone in particular in a public space is not as accurate since outdoor scenes vary and change so much.
The way to improve automated video analysis is by devising a software that can deal with this variability rather than treating as an inconvenience. One area that deals with large amounts of very variable data is genomics. The three billion DNA characters of the first human genome (the entire set of genetic data in a human) were sequenced in 2001, and since then, the production of this kind of genomic data has increased at an exponential rate. Given the amount of this data and the degree to which it can vary has led to vast amounts of money and resources being deployed to develop specialised software and computing facilities to handle it.
Thanks to this software, today it's possible for scientists to relatively easily access genome analysis services to study all sorts of things, including how to combat diseases and design personalised medical services, and even uncovering the mysteries of human history.
By investigating the mutations that have taken place over time, genomic analysis studies the evolution of genes. This can be compared to what visual surveillance is up against, which is the challenge of interpreting the evolution of a scene over time to spot and track moving pedestrians.
If we apply the same principles that are used in genomics to video surveillance, treating the images that make up a video as mutations, that we can solve the system's biggest challenge.
This practice is called vide-omics and it has already demonstrated its potential. For instance, one research group led by Jean-Christophe Nebel, associate professor in Pattern Recognition, Kingston University, has, for the first time, show that videos could be analysed even when captured by a freely moving camera. By identifying camera motions as mutations, they can be compensated so that a scene appears as if filmed by a fixed camera, Nebel explained.
At the same time, researchers at the University of Verona have demonstrated that image processing tasks can be encoded such a way that standard genomics tools could be exploited. This is particularly important since such an approach reduces significantly the cost and time of software development, Nebel said.
Combining this with our strategy could eventually deliver the visual surveillance revolution that was promised many years ago, he added. If the ‘vide-omics‘ principle were to be adopted, the coming decade could deliver much smarter cameras. In which case, we had better get used to being spotted on video far more often.