Drones and AI go hand in hand. Any machine that operates with a degree of autonomy needs to in some sense understand its surroundings. There are also obvious goals to achieve: a flight path, an obstacle to avoid, a point of interest to survey.

But aside from getting off the ground and the basic functions of today’s flying cameras, artificial intelligence is proving useful when it comes to the analysis of drone data. Drones can gather data from all kinds of sensors in all kind of scenarios, and AI is helping us make sense of it all.

Here are four examples of drones and AI coming together with spectacular results.

Measuring and Identifying Whales

Technology giant Intel has its fair share of investments in the drone industry, including Chinese manufacturer Yuneec and German aerial mobility company Volocopter.

Back in 2017, the company supported conservation work by Ocean Alliance from a software point of view, as part of the Parley For The Oceans program. Oceans Alliance has become well known for using drones – ‘SnotBots’ – to capture biological data from whale blows.

Ocean Alliance and Intel teamed up to collect and analyse whale data on a research expedition to Alaska.

Intel devised a machine learning platform capable of analyzing the live images being streamed from the drones. As a result, researchers were able to identify and catalog whales in real time.Previously, biological samples were required as part of a time-consuming effort to identify the whales and their sex. AI made it possible to measure and record certain details before the drones were even back on the boat, despite limited visibility at sea.

Counting Sheep

The idea of counting sheep might be enough to put some people to sleep, but for RIIS it’s another way to utilize aerial photography and AI.

Last year, as part of our ongoing efforts to explore the opportunities in drone data, we built the Sheep Counter application. It’s one of several drone apps we’ve worked on. Sheep Counter was born thanks to DJI’s Mobile SDK, which has enough functionality for iOS and Android developers to create drone-related applications without having to write loads of code from scratch.

But delving into DJI’s Mobile SDK was just the start. We wanted to explore ways that AI could increase the functionality of a drone. Inevitably that made analysing the data our key concern.

The Sheep Counter app uses object detection tools from TensorFlow – Google’s open-source machine learning platform – to identify and count individual sheep in a field. Sheep Counter allows drone pilots to plot a flight path around a field, import the drone images and stitch them together to provide a total headcount.

The app is a great example of how drones and AI can be used to quickly count objects. We anticipate that plenty of use cases, in animal husbandry and otherwise, will stem from this kind of approach.

Replacing The Need for Pilots

Computer vision is a much talked about area of artificial intelligence in the drone industry for obvious reasons.

For starters, several manufacturers already deploy advanced computer vision to stop drone pilots from getting into trouble: detecting an obstacle and autonomously figuring out a way around is a feature that comes with certain DJI drones, for example.

Last year California manufacturer Skydio took the technology to the next level, introducing a new drone to the market – the R1 – that completely removes the need for a pilot. It’s designed to film the user while following and avoiding obstacles autonomously. And it’s navigation capabilities go beyond anything offered by DJI.

Skydio has since made its Autonomy Engine open to developers to devise commercial applications.

Moving forward, this is an area of substantial R&D for obvious reasons: many of the more advanced drone applications, including urban deliveries and aerial mobility systems, will require further advances in computer vision and hazard awareness.

One interesting project highlighting the progress in this space is AlphaPilot, a drone racing competition with a difference. It’s organized by the Drone Racing League (DRL) and sponsored by Lockheed Martin.

The surface aim is to promote DRL’s new Artificial Intelligence Robotic Racing Circuit. But beyond that is a drive to research and design AI pilots capable of keeping up with their human-controlled counterparts.

Later this year nine teams will compete through a series of qualifying rounds for a shot at the $1m grand prize. There’s extra money on offer for any team that brings along an autonomous system that beats one of DRL’s professional racing pilots.

Protecting Architectural Heritage

With the devastating fire at Notre Dame cathedral in Paris making headlines around the world, it’s heartening to hear that drone technology was on the scene to support firefighters and prevent the blaze from causing further damage.

Now the emphasis turns to the rebuilding and restoration process.

Drones and AI are already being applied to similar projects. French startup Iconem is leading the way in this field. Founded by architect Yves Ubelmann, Iconem is setting new standards for the 3D modelling and documentation of important historical sites.

Many are threatened by looting, urbanization, mass tourism, conflict and climate change. Preserving them has therefore become an international priority. After all, these are sites where cultures emerged and civilisations started.

The aim is to preserve and protect global heritage, but the approach could not be more modern. Iconem combines drones with sophisticated modelling algorithms and cloud computing.

The result is highly-detailed and immersive models of major historical sites, that can be shared with researchers, restorers and the general public. Iconem has contributed to restoration projects, exhibitions, augmented reality experiences and more.

Malek Murison