Industrial drones are the new 'sensor network'

We usually think of aerial drones as consumer technology or used by the military to fly pilot-less missions. But there is an entire industry dedicated to using drones in industrial settings like mining, construction, and insurance.

The technology of industrial drones is fascinating and involves components such as ruggedized flight bodies, GPS, and LIDAR, in addition to cameras. These industrial drones combine data collection from a network of sensors with advanced image processing techniques and artificial intelligence.

So I could learn more, public relations ninja, Laura Hoang, introduced me to George Mathew, CEO of industrial drone supplier, Kespry. The company provides a complete enterprise platform for industrial drones that includes hardware, software, and rich data analytics.

Because drones are both cool and important, I invited George to take part in Episode 277 of the CXOTalk series of conversations with the world's top innovators.

During our talk, George explains drone tech and talks about topics like volumetric analysis in mining, LIDAR on drones, photogrammetry, and 3D orthomosaic image processing.

Watch the video embedded above and read the complete transcript. Below is an edited transcript of key points raised in the discussion.

What are the use cases for industrial drones?

They can fly seamlessly over industrial worksites into broader, commercial use cases where you can easily and reliably collect data. Drones can collect information in a meaningful, easy, cheap, and safe way. The drone becomes not only a method of data collection but also a way to transmit all kinds of new sensory-based input into analytics and models for industrial use cases.

One of the first industrial use cases that we started was measuring the volumetrics around a mining aggregate's operation. Mining in aggregates, you're digging things out of the ground. That's the primary product on which you're managing inventory. To understand how much material you have, you effectively have to be able to do a volumetric analysis of how much material you've pulled out of the ground.

Previously, the primary method to do that was taking survey grade equipment, whether precision laser-guided equipment or a GPS backpack, climb a stockpile and take 10, 15, 20 points of measurement. Then, go back to the home office and measure those volumes.

What's amazing about a drone now is you can fly over a worksite, a mine site, or an aggregates location, be able to take imagery of all of the stockpiles and material and convert those into three-dimensional models. Those models are hyper-accurate because you're taking 600,000, 700,000 points of measurement on a typical site and being able to get a level of accuracy on your inventory that was unprecedented. That's where we started in the industrial use cases that we're serving today by being able to do better and more accurate measurement in the mining space.

Also: 3D vision and AI are about to solve the biggest problem in construction

Since then, we've expanded into topological models for construction worksites around earthworks projects where John Deere is now reselling Kespry drones. We've expanded into the insurance and roofing world where we can do topological models of a roof and be able to assess the damage on a roof from weather events and, more recently, expanded into the energy sector.

What do drones add that is new?

Historically, there are two changes that drones, and other similar sensor-based technologies, introduced into the market. In effect, if you're getting information from a map, it's two dimensional. Drones now provide a level and a view in the third dimension: being able to understand elevation, being able to understand topology, being able to understand the volumetrics related to material that might be dug out of the ground. That capability of understanding the third dimension becomes quite relevant, particularly when you can do it in a super accurate way.

When we introduce precision GPS combined with data processing, what's incredible is we can create a topological analysis that's down to three centimeters of real space X, Y, and Z. Inaccurate, two-dimensional maps in your mining site are now realized as a fully available, three-dimensional, topology of that entire landscape as it changes over time. That's where we start to see the benefit of drones and added sensor-based input in this kind of industrial work.

What sensors do you include in the drones?

There's a visual sensor, a high-fidelity camera that can take imagery on the worksite itself. But, alongside that visual imagery, we have a one-dimensional forward-facing LIDAR.

The LIDAR, as a sensor, can be used for collision avoidance, obstacle detection, as well as a terrain map. We see use cases where, as the drone is autonomously flying on a worksite, you can have imagery autonomously taken, to see if there is an obstacle and avoid it in real time. That's possible because of sensor-based input.

Also: How 5G will impact the future of farming and John Deere's digital transformation

On the top, you see GPS. We introduced precision GPS into the Kespry product about a year ago. Now we can fuse the data that comes off additional sensors like the visual sensor, the gyrometer, the accelerometer, with the precision GPS to get topological analysis coordinated down to three centimeters of real space X, Y, and Z.

This is why I think the drone is a new sensor network that can start to apply these insights in a combined fashion that you, frankly, couldn't do most seamlessly before technology like this was brought together into the market.

What's the difference between industrial and commercial drones?

Kespry has introduced industrial capabilities into the market for ruggedized work areas like a mining location or construction location. You can potentially fly a consumer drone, but the challenge is that a consumer drone will have difficulty being able to be operated on a mine site that's 8,000 feet in altitude. It'll have difficulty in flying in 25-mile-an-hour winds. It has difficulty flying more than 15 minutes, and our drone today flies for almost 25-, 30-minutes and covers 150 acres when it flies at about 150-, 200-feet in the air.

And so, these qualities of industrialized work being accomplished with a drone needs a different type of product in the market than a consumer grade drone that has typically been in the market for quite a few years. This is where Kespry's focus has been in not only delivering that industrialized drone, but also the data processing, the applications that support the use cases that are necessary for the markets we serve.

Integration with enterprise systems must also be important?

That's right. If you're bringing a consumer drone in, you can potentially collect all that data, but then how do you process that? That's going to be manual, right? How do you make the insights possible? Well, you're still going to be going through a bit of a kludgy, broken experience. Then how do you get that data into an application that's consumable or a set of APIs that could be exposed to downstream applications? All of that tends to be laborious work that we help effectively prepackage and deliver to the customer base we're serving.

What kind of analytics do you use on the drone data?

One of the things that we care about primarily is integrating and owning the physical model of how a worksite and asset area is effectively understood from an analytical perspective. One of the key things that have been a differentiating factor in the market regarding automating this data collection and being able to generate analytics from it is a technique called photogrammetry. What I mean by photogrammetry is that you can take the 2D images that are collected off of a drone and, if you have the right overlap available of imagery and the right angle that you've captured that imagery, you can convert those 2D images into fully realized, three-dimensional models.

We started with that idea to, first and foremost, generate that 3D model, generate what's known as 3D orthomosaic, and then layer in all the aggregations, calculations, machine learning algorithms, artificial intelligence directly on top of the three-dimensional model. This is a natural progression of how industrial work continues to get accomplished because better decision-making is available; the physical model is easily exposed to a level of unprecedented accuracy with data and analytics.

Where is drone technology going?

Historically, you've taken sensors, flown over worksites, flown over commercial locations, and figured out how to manually get insights out of that information.

The market needs better automation, better data processing, better capabilities to get those insights off of the sensors; and, make automated sense out of them using machine learning, using artificial intelligence.

We've never built a solution with even a joystick. With autonomous capability, you punch in the coordinates to fly by casting a geofence and then drawing out the area with your fingers on an iPad. Hit the start button, and the drone autonomously takes off and does the work that it needs to do.

Also: Robots and drones: Coming soon to a construction site near you

We see this as natural automation and set of applications that support the data that's coming off the drone versus someone manually flying, which has been historically what the drone market has looked like.

It's not that the two are in violent conflict with each other. There are perfectly adequate needs to be able to fly manually. A good example is doing a bridge inspection, which is best-suited by a pilot manually flying a drone to collect that insight.

But, to look at shipping containers that might come to a port, or the dimensional analysis of a roof, or the operating efficiency of an entire solar farm, those things can be automated without pilots manually flying drones.

CXOTalk brings together the most world's top business and government leaders for in-depth conversations on digital disruption, AI, innovation, and related topics. Be sure to watch our many episodes! Thumbnail image Creative Commons from Pixabay.