VisionAR – How Computer Vision is Giving Drones Human-Like Eyesight

Dec 8, 2016

Reproduced from

Many U.S. industries, such as agriculture and industrial inspection, depend on the use of drones, also known as unmanned aerial vehicles (UAVs). The Federal Aviation Administration (FAA) estimates the “total hobbyist and commercial UAS (unmanned aerial systems) sales will rise from 2.5 million in 2016 to 7 million in 2020.” As drones become more popular, Iris Automation aims to satisfy the need for a safer and energy-efficient way to fly autonomously.

An Introduction to Iris Automation

CEO Alex Harmsen, recognizing a need for high-tech computer vision solution for flying drones, co-founded Iris Automation in 2015. Since its establishment, Harmsen’s company has raised more than $500k in seed funding from Bee Partners and Y Combinator, which has allowed the Vancouver-based business to open a second office in California and hire additional staff. Though the initial software did not require too much funding to get up and running, Harmsen says that the additional investments have been imperative for the growth and promotion of the product. With another round of funding currently in the works, Iris Automation plans to continue expanding by reaching out to new customers and the industry’s leading CV and AI engineers.


Industrial Drone Collision Avoidance

Due to inferior or nonexistent sense and avoid systems, a drone is required to remain within the visual line-of-sight of the pilot under current FAA regulations. Iris Automation is “finally unlocking beyond visual line-of-sight.” The company’s on-board CV technology “accurately tracks and avoids dynamic and static objects, both on the ground, and in the air at short-range and long-range,” Harmsen says.

Harmsen’s team understands that for UAVs to operate autonomously they must be able to see the world as humans do. Unlike traditional radar, Iris Automation’s CV allows UAVs to see and track objects in real time and have complete situational awareness. For example, an autonomous plane powered by Iris Automation software can tell the difference between a bird and a balloon and track the movement of either in order to decide the best action to take to avoid a collision.

While self-driving cars function with similar technology, the space that cars operate within is much more structured than that of an aircraft. Self-driving cars can rely on roads, maps, and traffic patterns. A car, which only operates on a single plane, may only avoid objects by moving side to side or front to back, but a drone has the ability to maneuver up and down, as well. While there are more obstacles on the ground, there is more space in the air that must be scanned and monitored.

Though it is much more likely for a car to hit something on the ground than it is for a drone to hit something in the sky, the impact is greater for the aircraft. Due to speed and elevation, flying drone collisions are much more detrimental. Iris Automation’s software aims to give drones complete situational awareness with a 140-degree field of view allowing drones recognition of potential collisions sooner in order to decide the best evasive action.

The Nuts and Bolts of Iris Automation’s Software

The idea behind Iris Automation’s Industrial Drone Collision Avoidance system is quite literally driven by its high-tech CV and AI, but Harmsen sought to create a superior product to support the technology. The company’s passive sensing methods allow for a low-power product that will not drain the drone’s battery life. Additionally, they have created a small, lightweight modular design that “fits on almost any drone, custom and commercial.”

Most of Iris Automation’s current customers are interested only in the software, however, Iris Automation has developed a hardware piece for drones that require the necessary embedded GPU.

Iris Automation has created a scalable product that is low-cost (approximately 10 percent of the cost of a drone).

The Direction Drones are Heading

Harmsen believes “everything is going in the right direction” for his company. After a new FAA ruling went into effect earlier this year, it is much easier for businesses in the U.S. to operate drones. According to a recent Insurance Journal article, the FAA projects that “drone manufacturers will turn out a fleet of about 542,500 small commercial drones that will serve industries and be permitted to operate in nation’s airspace” within the next five years. Additionally, U.S. Secretary of Transportation Anthony Foxx has reported that industry research projects “the commercial drone industry could bring in $82 billion for the U.S. economy as well as 100,000 new jobs by 2025.”

Harmsen says, “It makes sense the government wants to have safe systems. We’re happy to make them safe.”

Once Iris Automation can prove it has created a fail-safe CV system, robot pilots will be just as trustworthy and safe as human pilots. The ultimate goal for Harmsen is that drones will be able to operate completely autonomously and out of the visual line-of-sight of human pilots. With advanced CV tech, Iris Automation appropriately claims they have “the future in sight.”