The video analytics market is all set to snowball in the coming years. According to a report from Memoori, significant advances in semiconductor architecture that have enabled better use of deep learning and machine learning algorithms could drive this market.
The global traffic management market size is expected to grow from US$ 30.6 billion in 2019 to $57.9 billion by 2024, at CAGR of 13.6 percent, according to a report from MarketsandMarkets. The increased use of surveillance technology in traffic management is set to boost the use of video analytics in this segment.
In this article, we take a look at some of the latest features in video analytics for traffic management and surveillance, their benefits, best practices, and hardware requirements.
Video analytics has a major role to play in the management of traffic, and enforcement of road rules for passenger and pedestrian safety. Traffic surveillance cameras have become ubiquitous in most major cities and highways worldwide, and footage from these are the perfect raw material for today's intelligent video analytic solutions.
Over the last two years, rapid developments in this segment have led to more and more sophisticated algorithms and efficient applications. Algorithms have evolved to automate many processes that can help law enforcement agencies to ensure vehicles adhere to regulations and immediately know if there are accidents or other mishaps.
Average speed ANPR
According to Ranjith Parakkal, CEO of
Uncanny Vision, a critical feature that has become quite useful to law enforcement in the recent times is Average Speed ANPR.
"Earlier, detection of overs-speeding vehicles was based on the recording of spot speed that speed radars would capture," Parakkal said. "But this is a system that drivers can easily beat once they know where the speed radars are installed. You can slow down your vehicle close to the speed radar and accelerate after the device. Average Speed ANPR works using road signs/gantries or speed detectors with cameras that are placed at short intervals like every three kilometers or so. As a vehicle drives through this stretch, every camera captures its number plate. The system then calculates the time the vehicle took to cover the distance between two cameras, to find an average speed. This is a perfectly accurate and fool-proof system with no possibility for error."
However, this solution relies heavily on the accuracy of the number plate recognition. If the ALPR system is not reliable, the overall performance of the system decreases. Some experts have also pointed out that this system is a bit more challenging to enforce a law compared to peak speed detection radars since the latter offers a straightforward result that cannot be questioned.
"The problem is legislation," explained Jermain Santoya, Industry Marketing Manager at
Genetec. "The reason spot detection is used so much now is, at least in North America, you need to prove that a person was speeding to issue a ticket. With radar, you have a way of knowing that the person was going over the speed limit. However, with an analytic solution, it is a little bit blurrier, with so many different factors and sensors. We will then have to reach a point where the legislation is comfortable to say this is enough proof to issue a ticket to someone and punish them for speeding."
In short, it's about where we are in terms of technology and where we are in terms of its acceptance.
Mobile phone usage detection
Video analytics based on footage from surveillance cameras installed on highways and other roads are also useful in detecting if the person driving a vehicle is holding a mobile phone and using it. Many countries have regulations that prohibit the use of mobile phones while driving, although using them with handsfree or the speaker system is allowed in some places.
Parakkal said that this solution is a bit tricky as the glare of the windshield during the daytime could adversely affect the system's accuracy. However, it would work more efficiently in the night, with IR cameras that can see inside the car. The adoption of this feature is not yet widespread, as it is rather new.
ALPR-based parking
Before, when a vehicle drove up to a boom barrier, access to the parking lot would be granted using a contact-based or contactless RFID readers. But now license plate-based access control is increasingly becoming popular. License plates are unique identities, and if the system can recognize them, there is no need for another ID. This can work in paid parking spaces as well if the license plate is attached to digital wallets.
"The main advantage is that you can do away with the hassles of RFID tags and their distribution," Parakkal said. "But the benefits don't end there. This system can help in visitor management by providing access to a visitor's car after its license number is entered into the system."
Seat-belts and other violations
Seatbelts are integral to road safety, but there are many instances where drivers and passengers choose not to wear them. Like the mobile phone-usage detection feature, this also requires cameras to capture visuals of a vehicle's interior, and hence the glare from sunlight could hurt accuracy levels. However, such algorithms are evolving quickly, and we may soon see better results in the coming years.
Istik Kattan, CEO of
Agent Video Intelligence, pointed out that these features are relatively new, having entered the market in the last two years or so. According to him, apart from what is already mentioned, AI-based vehicle classification ( ability to make a distinction between various types of groups of vehicles and derive alerts / statistical analysis tied to specific vehicle classes) and traffic violation detection (double parking, parking in prohibited locations, passenger vehicle in public transport lane, turning from the wrong lane, etc.) are also becoming popular now.