Count on deep intelligence
Stuart Bettle, Video Product Marketing Manager EMEA for Tyco, the security products division of Johnson Controls, showcases the new VideoEdge Deep Intelligence NVR, Tyco's first step towards developing machine learning solutions that incorporate advanced forms of video analytics and artificial intelligence.
People-counting has come a long way from when businesses had to employ staff to stand at doorways with handheld clickers. Airports, train stations, night clubs and retails stores, as well as museums and tourist attractions, are now just a few of the environments where video analytics is helping provide valuable management information on the number of visitors who enter a premises within a defined time period.
In a highly competitive environment, people-counting can help retailers to measure the effectiveness of marketing campaigns in attracting customers to visit their stores or gain an understanding as to why one store is performing better or worse than others. It can also assist with the scheduling of staffing levels to ensure costs are kept to a minimum during quiet periods, while maximising the customer experience during busy periods.
Non-profit or government funded organisations, such as art galleries and museums, are also increasingly relying on video analytics to provide data on visitor numbers in support of grant applications, while leisure facilities are turning to people-counting technology to verify the number of visitors to ensure they do not breach licensing regulations, says Stuart Bettle, Video Product Marketing Manager EMEA for Tyco.
There is certainly no shortage of manufacturers offering video analytics based people-counting solutions. Some of these operate on the edge, ie, onboard a camera, while others need the processing power of a server. People-counting accuracy is likely to vary, depending on the ability of each type of analytics to cope with varying environmental conditions or the number of people who may cross a threshold at the same time.
Machine learning occurs when a computer utilises complex algorithms that process data to make predictions. The algorithms are devised by engineers in order to guide the device as it learns about an environment or situation. Visual and numerical information stored as metadata is gathered and then analysed by the computer in massive data sets via an artificial neural network (ANN).
The advantage of deep learning neural networks compared to other types of artificial neural networks is that they are able to learn the key features of patterns automatically.
Tyco, the security products division of Johnson Controls, has launched an eight-channel VideoEdge Deep Intelligence Network Video Recorder (NVR) that utilises machine learning techniques with the help of a powerful graphics processing unit (GPU). This optimises the ability of the NVR to display highly accurate video intelligence compared to standard methods of video analytics.
Accurate people-counting and tracking is achievable from images captured by an overhead camera as the VideoEdge Deep Intelligence NVR is able to more effectively distinguish between humans and objects in the field of view. This is because it is supplied 'pre-trained' with thousands of images to compare against and, as such, will work straight out of the box for the majority of applications.
Greater situational awareness
When combined with the full analytics suite available on the VideoEdge Deep Intelligence NVR, such as crowd formation, loitering and perimeter protection, together with the command-and-control capabilities of the victor Video Management System, users are now able to identify opportunities to improve productivity, as well as achieve greater security and situational awareness.
E-mail BTS-EMEA-TycoSales@jci.com for more information about the VideoEdge Deep Intelligence NVR.