Drones are big business, with Goldman Sachs recently forecasting the market will grow to $100 billion by 2020. These stellar forecasts are fuelled partly by the fact that the drone provides a platform for a host of emerging technologies that are all converging to enable applications never before possible. Computer vision, deep learning, autonomous transport and AI, coupled with the slick consumer electronics supply chain, have resulted in devices that are truly pushing the boundaries of technology while being affordable to consumers and businesses alike.
As well as becoming hugely popular with hobbyists, drones are finding many applications in business, from construction to real estate. It’s no surprise that the commercial sector is turning to drones, simply because of the cost: for example, a drone could be used for regular inspections of a pipeline, a task that would previously have required the use of a helicopter. The drone can handle the same amount of work, surveying up to 150 miles of pipeline per day, but its cost is a small fraction of the $2,500 per hour expense of hiring a helicopter and crew.
The New York Power Authority found that an infrastructure inspection by drone cost less than a tenth of the same test performed by a helicopter or boat. And there are many other inspection applications where drones can save huge amount of money, such as agriculture.
AI and autonomy
To help make such applications a reality, today’s drones are becoming more and more sophisticated in terms of Artificial Intelligence (AI) and autonomous capabilities. On-board gyroscopes, accelerometers and air pressure sensors provide accurate data on position, velocity and acceleration, enabling the drone’s processor to follow a pre-programmed path, or to land safely without human intervention.
AI has taken this a step further – for example, researchers at Carnegie Mellon University recently showed how a drone could teach itself to fly through complicated courses, avoid obstacles, and navigate narrow gaps.
The state of the art in consumer drones
So what do today’s latest drones look like?
Let’s use the example of the Mavic 2 Pro, from DJI, one of the most popular drone vendors. Announced a few weeks ago, this has sophisticated autonomous capabilities, such as flying in a circle around a subject to create a time-lapse video. It’s also DJI’s first consumer drone that has camera powered obstacle-detecting sensors on all sides, to help it prevent collisions.
But the Mavic 2 Pro isn’t just about clever features. More importantly for most buyers is the quality of the camera. Last year, DJI reportedly acquired a majority stake in the iconic Swedish camera company Hasselblad. Through this investment, DJI designed and incorporated a Hasselblad camera into the Mavic 2 Pro that is so good that it can compete with DSLRs in terms of quality, and which enables the drone to capture stunning video and still images. This means the Mavic 2 Pro can now enable industries such as film-making and farming to dramatically cut the costs of applications, which would otherwise require extremely expensive equipment.
Under the hood
With sophisticated flying and image-capturing capabilities, today’s top-end drones have some serious on-board processing. In fact, the best example of how all this technology can be put into practice is DJI, which has been highly successful in creating and expanding the drone market, winning more than 80% of the consumer sector and more than 60% of the professional segment.
DJI has achieved this with robust, high-quality designs, that are simple to use and affordable. It has developed new products at a relentless pace, and beaten the competition in how fast it can release updates and fix problems. It has also taken in-house the design of complex components, building unrivalled expertise in hardware and software, thus enabling it to drive down costs while improving features and performance.
Already being deployed in drones from DJI and others, CEVA vision processors and AI accelerators have proved an excellent fit for the signal processing and machine learning requirements of these high-performance autonomous flying machines.
The drone market has come a long way over the last few years, but in many ways, it feels like it’s just getting started, with new applications opening up. The quality of photography is only going to improve and, in parallel, processors will get more powerful and AI will get smarter. This means that drones will get even better at autonomous flying, image recognition, and decision making, resulting in them being able to deliver ever more sophisticated capabilities and services.
More legislation and regulation seems inevitable, as governments seek to control the fast-growing number of drones and to maintain safety in the air.
But the potential of the drone is enormous – it is already a disruptive technology and, whatever the application, looks certain to become an increasingly influential force in the years ahead.
The Article was published in Embedded.com
You might also like
More from Deep Learning
In recent times, the need for real-time decision making, reduced data throughput, and privacy concerns, has moved a substantial portion …