Insights

|

17 Apr 2020

Edge intelligence and LPWA: Improving cost, coverage, and confidentiality of video surveillance solutions

On the road, at home, and at work, smart surveillance cameras are finding new applications as advanced IoT data sensors.

Thermal image of public transportation passengers in a city

If a picture is worth a thousand words, high-definition video must be worth much, much more. It’s what’s made the technology so popular in such a broad range of applications, including online face-to-face communication, entertainment, domestic surveillance, access control, and more. But sometimes all you want is just one word: a yes, a no, or some other single insight gleaned from your camera. In a growing number of use cases in which surveillance cameras are merely a means to gain valuable insights, not video footage, edge intelligence and low power wide area communication are delivering unmatched performance while improving cost, coverage, and user confidentiality.

Smart traffic applications illustrate this well. Take car license plate recognition: Rather than uploading the entire video feed to the cloud for processing, artificial intelligence (AI) algorithms on the camera can, instead, simply extract the plate number and transmit it to the backend. AI-enabled cameras can also be used to identify drivers talking on their cellphones while at the wheel. At intersections, surveillance cameras are already monitoring vehicles and feeding municipal traffic management systems to control lights. Because the images never leave the camera, they privacy of those that happen to pass through the camera’s field of vision can be protected.

In the domestic setting, smart surveillance cameras can replace conventional sensors used to detect fire, smoke, or gas. They can use facial recognition to detect intruders, or analyze images to catch other incidents such as falls, with the additional benefit of making it possible to switch to a higher bandwidth service to remotely check in on the situation in high definition video. And when deployed in a mobile setting, battery-powered cameras featuring a GPS receiver can set off geo-tagged alerts to speed up the intervention of rapid response teams.

In shops, malls, and other commercial venues, smart surveillance cameras offer a powerful tool to monitor the flows and the density of people, again safeguarding privacy by processing the images on the cameras rather than in the cloud. Gathered data can help shop managers study customer behavior. Of particular relevance during the ongoing covid-19 pandemic is their ability to ensure that occupancy restrictions and social distancing measures are respected at all times.

And by observing frequencies outside the visual range, surveillance cameras can overcome some of the limitations of human sight. Thermal imagery, for instance, is less affected by fog, glare, and adverse meteorological conditions. This makes it ideal for a number of smart city use cases. At the same time, thermal imagery protects privacy by essentially anonymizing those observed. Hyperspectral imagery, which divides the light spectrum into dozens of narrow frequency bands (compared to the three used by our eyes), opens up entirely new avenues in image analysis by revealing features that otherwise remain elusive. Use cases include quality monitoring in industrial and agricultural production and processing lines.

The ultimate IoT sensor

By analyzing the video stream at the edge of the network – e.g. on (or near) the camera, typically using trained machine learning algorithms, the surveillance cameras essentially become smart IoT sensors that communicate only actionable data. Compared to traditional IoT sensors that only observe a single physical property, smart video cameras capture a wealth of data that, using the right algorithms, they can extract.

Compared to conventional surveillance cameras, they cut their communication needs by reducing the amount of data transmitted. While live-streaming HD video requires up to 25 Mbps of bandwidth, connected surveillance cameras that use edge intelligence to extract insights from image data can require as little as just a few bytes of data to send out an alert in the case of an incident. It’s quicker and more robust as well: Performing the analytics at the edge can be done in a fraction of the time required to transfer to data to a cloud server and back.

Thanks to advances in storage media, it is now even becoming possible to get the best of both worlds. Many sensitive use cases require or might benefit from the possibility to access video footage. The latest generation of microSD cards, with up to a terabyte of data storage, can record three months’ worth of video footage  right on the device, giving service providers a means to reconstruct the cause of observed incidents or services outages.

Cost, coverage, and longevity

The reduced bandwidth needs that characterize the aforementioned applications make smart surveillance cameras ideal candidates for low power wide area (LPWA) cellular communication technology, such as NB-IoT or LTE-M, which offer less expensive communication modules and data packages, as well as expanded coverage into otherwise hard-to-reach locations such as deep inside buildings or underground. And because data transmission is typically far more power-hungry than on-device image processing using artificial intelligence algorithms, leveraging power-optimized LPWA technology means that battery-powered smart surveillance cameras can operate as IoT sensors for months at a time.

 

Diego Grassi

Senior Manager Application Marketing, Industrial Market Development, u-blox

You might also be interested in