Nowhere to hide: Building safe cities with technology enablers and AI
Governments are ramping up camera coverage in cities to stop "shadow maps" of crime hotspots from emerging and are using big-data driven AI like computer vision to cut crime.
In the popular sci-fi crime drama Person of Interest, criminals are equipped with a shadow map of New York to avoid surveillance cameras, creating breeding grounds for crime. Shift to the real world, and governments are ramping up camera coverage in cities to stop these hotspots from emerging and are using big-data driven tech like computer vision to cut crime.
Security cameras are everywhere – streets, squares, parks, shopping malls, airports, train stations. But, there are plenty of obstacles to optimizing their use. One prominent issue is that authorities can’t access privately run cameras. For example, the UK is home to 6 million cameras, but only 70,000 are run by local authorities, meaning they need permission to see private footage.
The second is the conflicting goals of privacy and public safety. Many countries ban cameras on residential roads to protect privacy, meaning that undisputable visual evidence is lacking if a crime occurs. A crucial part of building a safe city, though, is reducing the number of areas that can make up a shadow map.
Enter AI – the new kid on the block for public safety.
In one episode of Person of Interest, a super AI named Machine identifies a government employee as a possible terrorist by associating 18 gas station receipts with other data sets. Machine discovers a regular pattern of visits to a particular gas station on the third Thursday of every second month, despite the antagonist having filled up with gas the day before. On three occasions, an SUV registered to the wife of an executive would pull up two hours later – Machine found that the executive had paid for plane tickets for a suspected terrorist back in 1994, forming a possible link. The corrupt government official was arrested and tried for his crimes. Though TV shows of this ilk don’t shy away for stretching credibility or the occasional glaring plot hole, the above scenario is no longer implausible thanks to big data and analytics technologies, both of which will result in a qualitative leap in the power of AI.
When constructing safe cities, different municipal departments and telcos will be able to pool resources and build integrated systems that combine street lamps, mini base stations, and surveillance cameras, so that all areas with street lighting can be placed under surveillance. It will be possible to beam footage from front-line police officers equipped with wearable cameras live to control centers on wireless networks.
In the fully connected cities of the near future broadband is fast and ubiquitous, allowing public spaces to be covered by surveillance equipment, with unified platforms incorporating information from a range of sources, including environmental monitoring equipment, road surveillance cameras, neighborhood and home security systems, and network information security surveillance. Control and dispatch centers will use this information to help carry out unified surveillance, safety management, and dispatch of public safety resources.
However, it will be much more difficult to quickly process such massive amounts of data to discover crime-solving leads. For example, a medium-sized city in China produces over 300 PB of video surveillance data a year, which requires an incalculable amount of resources to analyze given current scanning methods.
Doing so will require intelligent recognition technology built into cameras themselves, with front-end recognition capabilities carrying out facial recognition of everyone captured on video and analyzing specific behaviors to perform real-time crime prevention. Currently, a common police crime prevention tactic is to target people with criminal records, but this kind of presumptive method is not especially effective at stopping crime from taking place.
As computer vision advances, surveillance cameras will be able to alert police to crimes taking place based on people’s behavior, not just their identity. A surveillance system could for example help prevent a violent crime from occurring by immediately alerting nearby police officers when a weapon is detected, whether that person has a criminal record or not.
To achieve this requires computers to perform deeper and more detailed analysis of massive amounts of surveillance video footage uploaded to cloud. It won’t be enough to just identify and index people and scenes captured on video; it will need AI that can perform association and reasoning. By analyzing people’s behavior in video footage, and drawing on other government data such as identity, economic status, and circle of acquaintances, AI could quickly detect indications of crimes and predict potential criminal activity, just like Machine on Person of Interest.
In China, the first breakthrough in crime-fighting computer vision occurred with the vehicle recognition technology Yitu, a video-image car recognition system. Police in Suzhou, where the system was first rolled out, used Yitu to successfully apprehend criminals who stole over 100,000 yuan by identifying and filtering the getaway car model on surveillance – they caught the thieves in less than 10 minutes.
In scope of application, facial recognition outshines object detection. In the Hollywood blockbuster Mission Impossible – Ghost Protocol, an agent uses a facial recognition system to identify enemy agents in a train station. Today, that kind of technology is already available, with systems able to carry out static face comparison and recognition using databases containing hundreds of millions of faces. In China, Xiamen Police use this tech to reduce petty theft at the main train station by extracting still portraits of suspected pick pockets from video footage and running them through a static facial portrait system, quickly identifying names, ID card numbers, and the times and locations these thieves have previously appeared. The police then know when and where to patrol.
Since bringing Yitu’s computer vision technology into its ICT arsenal, Huawei has incorporated these AI capabilities to complement and enhance its Safe City solution with advanced computer vision tech.
In Maslow’s hierarchy of needs, safety is near the base of the needs pyramid, supporting self- esteem and self-realization, key factors that help advance humanity.
It might not yet be possible to build a city with no shadow areas, but advancements in AI and increasing digitization mean it won’t be long before we can use this technology to deduce criminals’ intentions and the tricks of the trade with the crime-fighting AI.
As nations begin achieving a level of ICT maturity that moves them into the stage of Augmented Innovation, enabling technologies like cloud, big data, IoT coupled with AI will mean that our world will be safer as well as smarter.