How drone and robot video can combine to reshape surveillance

Both drones and robots are useful tools for handling video surveillance, but the revolutionary tactics of merging the two technologies is creating a whole new strategy for across-the-board security.

Drones are appealing because they can fly high—quickly—and get an overview of a facility; something that was once only possible from a stationary camera in a high position. Their multiple rotors and lightweight batteries mean they can move and hover autonomously, which has certainly changed the game of security fields, as well.
Yet, there’s still a need for robots at ground-level because they have the smarts and battery life necessary for long surveillance periods. They can also carry more cameras and sensors than a drone. So, you can see why using the technology together is a smart move.

Teaming up

One eye in the sky was never going to cut it. Any surveillance team needs a range of information sources pulled into a simple user-interface that can quickly be interpreted—a holistic view.
So, while a drone can take an overhead shot like never before, it can’t necessarily interpret every on-the-ground detail. By contrast, surveillance robots can get up close, sense and interpret the activity of a given scene. These sorts of robots can also access online information to fill in any holes. Scientists continue to tinker with this, trying to arm robots with a greater ability to perceive what’s taken place and draw logical conclusions.
For now, this is still a team effort—a belief that Cornell University in the US holds close. The institution is currently testing the effectiveness of surveillance fleets , which combine cameras on both grounded robots and flying drones. The idea is that, by linking these machines to each other, they can piece together a more comprehensive surveillance picture—again, without the need for people at the wheel.

Making drones like robots

At a glance, the AI in robots is much closer to humans, which makes sense given that robots rely on learning algorithms . Drones can look and act like something much more akin to a science fiction movie. So, for robots and drones to work better together, the next logical step seems to be broadening drone capabilities to meet those of robots.

For example, Harvard researchers have revised their revolutionary ‘bee’ drone to enhance its capabilities. Once upon a time, it could only stick to walls. These days, it can fly, dive into water and go right back up into the air afterwards. While it might not be ready to interpret the way a robot can, you can imagine how valuable its size might be in previously inaccessible areas.

This type of aerial surveillance could be especially handy for law enforcement operations such as terrorist and hostage situations, assessing bomb threats, search and rescue missions or photographing crime scenes. For example, New South Wales Police has been using similar drones to survey large areas with limited access, such as dense bushland. Machines such as the DJI Inspire are helping units produce valuable high-definition imagery, given they are sturdy enough to fly high and at significant speeds. While certainly bigger than the bee drone, many police models are equally efficient and easy to deploy.

Making robots a little like drones

The argument for on-the-ground robots has always been around reliability. These robots typically have versatility and sturdiness, which means they can conduct prolonged tours of an area. However, now some technology firms are building robots that can fly and are able to last much longer in the air.

For example, the Parc robot (which is technically a rotor drone) can be set to fly indefinitely and at a specific altitude . More like an on-the-ground robot, it also carries a high-resolution camera capable of producing infrared footage for night vision. This is an example where the technologies have merged to improve what we already have on-hand.

Cutting down distraction

A significant challenge for surveillance technology overall, is that it can’t tell human operators about problems it sees. A robot can detect motion, sure, but this isn’t always accurate. In one example, surveillance manufacturer Avitas is trying to overcome this by building a new line of cameras equipped with deep neural network AI—developed by Nvidia—which helps robots detect when something isn’t right to send real-time alerts.

For instance, it might pick up when a person returns to the same spot multiple times within a short period. These new cameras can spot this behaviour, where a human operator monitoring a feed might not have noticed. Other similar systems count on deep neural learning to recognise patterns in data . This is especially good for image processing. In fact, it’s said that in some cases, deep learning can perform image recognition better than a person could.

A more intelligent future

AI is certainly making surveillance smarter. Where stationary cameras and sensors were once sufficient, AI is showing us that more perceptive robots and drones will make it significantly harder for criminals to succeed. Especially as surveillance machines become more independent.

For example, experts can train a surveillance vehicle to recognise or follow a particular object and at the same time, avoid obstacles in its path. In addition, AI development is also improving the control centre, with voice command likely to revolutionise the speed and accuracy of how we use both flying and on-the-ground machines.

What was once the domain of sci-fi flicks like Blade Runner, has turned into an exciting reality for the surveillance industry.

‘How to build a security plan for the future rather than the past’ will be one of the many topics discussed at this year’s ASIAL Security Conference. For more details and to secure your pass, please visit our conference page.