printer friendly version

Paving a path from LPR to AI

In the licence plate recognition (LPR) landscape, Visec is a name well known for having spearheaded the South African market. Several years ago, the company spun out its LPR services from its then core offering being video monitoring software, and under a new corporate structure it is now looking to evolve further in terms of both its technologies and its business model.

Visec was founded by South African Gary Scagell and American Alex Bordbar in 2002 and, in March this year, was sold to a consortium of new shareholders. With the ambition of taking the business to the next level, Visec employed chief strategist Jason Berry. “Initially I had a three-month contract to advise on the direction on the company, but I was so excited about where things could go that I was pleased to accept a longer term appointment,” Berry explains.

The foundation for the next stage of growth is Visec Cloud, a platform to provide services including the collection of licence plate information is taken from network-connected video cameras and automatically cross-referenced with a database containing vehicles of interest (VoI). Berry says what intrigues him most about this is how standard technology can be used in a highly effective way to fight crime, and the fact that this policing can happen at a community level and strengthens the bond to an underfunded police force, and further strengthen the relationship between SAPS and the community.

The integrity of the VoI database is key to the success of the Visec Cloud platform, since false alerts could lead to disastrous consequences. “The worst case scenario would be if a vehicle was registered incorrectly as a VoI, for example if it had in fact already been recovered or its details were recorded incorrectly, and armed police or security officers swooped on the vehicle with deadly force. Think of what that would do to the reputation of all parties, including SAPS, municipal law enforcement, the security company, or even Visec itself,” he poses.

For this reason the company provides the systems and support for the independent group of passionate members of the SLDC (SAPS LPR Data Custodians) who do the outstanding work in authenticating changes to this live database and that it contains the correct and complete information, and can only be modified or accessed by authorised persons.

The SLDC also do the background checks and authentication of users on the system. As a result, Visec Cloud is trusted by approved users who include the likes of SAPS members, the Hawks, the National Prosecuting Authority, counter-intelligence, environmental bodies, anti-poaching, city councils, metro police, law enforcement, approved neighbourhood watch operations managers, community policing forums, corporate market, large property owners/operators, border control, cash-in-transit and car rental risk investigators.

Learning on the fly

The next technological push for Visec Cloud is in the area of artificial intelligence (AI). To understand the potential this holds, Berry poses the all-too-common scenario of a vehicle being reported as stolen. Ideally the first step would be for the system to simply analyse camera footage to find the licence plate number and pinpoint its location, but the reality is there are people out there who have fake licence plates to avoid getting speeding fines, and if one of them happens to use that same number the system will throw up a false positive.

Visec’s AI can filter out these false positives by profiling the stolen vehicle’s history and determining by extrapolation whether, during the time since the theft, it could possibly have travelled the required distance to where the fake plate was spotted – if not, that particular detection can be dismissed as false. It can also provide a set of analytical functions to find the right car, and done at such a speed and manner that is nearly impossible by a team of control room operators.

Visec has also partnered with the artificial neural network (ANN) power house, Deep Data, a South African company whose founder, Jasper Horrell, is the brains behind big dataset handling at the Square Kilometre Array (SKA). Visec has also partnered with the Swedish company Irisity, which specialises in behavioural analytics. The result is the ability to develop AI features (which are currently in various stages of development) that can build a full profile of a vehicle based on its licence plate number, location and detail history, and identifying unique and changing characteristics of the vehicle. “This technology is only going to get better and better over time, since an ANN is essentially an ever-learning system,” Berry forecasts.

“Currently we’re reading between 1.2 and 1.5 million plates a day with a history going back five years, so we have a large dataset to train the system on. Visec’s focus is to continue to build the cloud-based platform, whereby third-party sensor-based info such as camera feeds can be analysed. At the moment, it’s LPR data, and soon we will release a full vehicle analytics solution. We are also currently testing the facial recognition capabilities of our platform and related mobile apps as we seek to expand our offering in the near future.”

A tangible example of how this technology is bearing fruit is a project Visec is working on with Cape Nature, to deal with the problem of perlemoen poaching. In collaboration with Irisity, it has demonstrated that an early warning system can be developed to detect suspicious activity such as divers hanging around on the beach for no reason, and even boats out at sea matching the expected behavioural traits.

Growing the business model

Currently Visec donates its services to SAPS without charge and has done so for years so as to prove the model and build the effective system, and it offers a licensing model of R350 per month, per camera for neighbourhood watches, discounts for other community based organisations, and R695 for a commercial operation. “So far it’s been all about getting the correct LPR cameras deployed correctly, but we are working to develop more commercial models around utilisation of the dataset – within the framework of South African law, of course.

“At the moment we track several thousands of lanes using these LPR enabled cameras, and this is growing rapidly. We are looking to at least double that every year for the next few years, and are on track to do so. The way we will go about doing that is to start at the grassroots, neighbourhood level to provide awareness, and provide mechanisms to empower the vast network of service providers in the industry,” Berry concludes.

Share this article:

Further reading: