Living in Cattle country, I’ve long been interested in things like using drones to check cattle (especially on those cold winter days when you can often find yourself out in 6+ feet of snow and temperatures as low as 50+ below. Having a good drone system and knowing how to use it can save time, trouble and even lives.
However, checking cattle with drones does have its downfalls. First, it’s hard to see the condition of cattle from the air. So if you’ve got a cow or steer with a snake bite in the middle of summer, it’s hard to pick up from the air on most of the drones I’ve seen. Second, in bad weather, when you would rather use a drone, the signal quality and fly-ability of a drone is often an issue. You can’t fly most drones in extremely low temperatures or high winds, heavy snow or rain, or darkness.
One last major issue is the difficulty in identifying a particular head. If you’ve been doctoring a cow and you need to know how she’s doing, it just damn hard to identify her from a drone. Even if you can I.D her, you may not be able to determine her condition from a drone.
Some of these issues can be addressed by technology. Some are more difficult. Such as weather. While we can build all-weather aircraft like those used in storm chasing, the required aerodynamics of such a bird simply goes against most of the other requires for a drone. All-weather type craft need to be able to penetrate the air so sleek fast designs are better for severe weather. Also, the heavier craft is less buffeted about by high winds that usually accompany severe weather. However, while many ranchers are indeed pilots, most couldn’t fly something like an F4J Phantom. J3 Cubs are more suited to both their runways and abilities.
Having a drone that flew 200 Mph or was heavy and needed long runways of smooth pavement simply won’t due in the ranching environment. What is needed there is a simple to pilot, simple to use craft with VTOL (Vertical Take-Off and Landing). VTOLs have their issues though. First, quadcopters and helicopters are less energy efficient than a fixed-wing craft. So range on a large ranch can often be an issue. This can be overcome with more expensive hardware. But with that comes more complexity and much larger price tags.
What is needed is a small craft capable of covering the kinds of large ranches you find in Montana, Nebraska, the Dakotas, Texas, and Oklahoma. Where a ranch could span hundreds of thousands of acres. This craft should be simple to operate. Should have a long-range transceiver and video equipment, and be able to transition from VTOL to high speed forward flight and back again. This would allow the craft to take off and land vertically while giving higher flight speeds while in transit to a large pasture. Once there, it should be capable of slowing to allow for better viewing of the cattle.
I believe AI can help solve some of these issues. First, a combination of vehicle design and AI flight stabilization can take care of much of the severe weather flying issues. Using AI to correct the flight path and react almost as fast as the wind, could make severe weather flying a non-issue. Insulated and heated battery compartments could help in improving efficiency in cold weather while improving battery life.
Using AI to stabilize the video and providing automatic tracking of cattle on the ground would be a great added feature. Furthermore, AI may be able to identify a particular head. Perhaps, even tell if a cow or steer is moving in an atypical manner.
I’ve put a lot of thought into this. Some of the challenges for such a system would be that identifying cattle from an image recognition system is almost impossible. To identify a cow you need 3D data and views that may be hard to get with a cow in the middle of a bunch.
Ear tags might help but you’ll have the same issue as most ear tags are pinned to the front of the ear and can be seen best when looking the cow in the face. Getting cattlemen to change this would be difficult as there will be many times when the rancher will simply have to be there! And if the tag is on the backside of the ear, he’s not going to be able to see it!
Adding another marker to the backside of the ear may be an option but your system had better work and work well if you’re going to have to round up all those cattle and re-tag them all. Also, any I.D. tags will need to be marked with some type of robust marking that can handle errors in reading easily. As tags often become bent, broken, covered in dirt and manure and even lost.
That brings up another issue. Anything you have to place on the cattle must be dirt cheap and easily replaced. This is why some type of ear tag might be the best route to identifying cattle from a drone. I’ve toyed with some ideas even writing a photo image recognizer for cattle a few years ago. It was written in Octave after taking Andrew Ng AI class on Coursera.
I am very interested in the ability to identify cattle from a drone. I believe that when cattlemen can pick out a single head from the air and track its position, how close it is to the rest of the heard, etc… That drone use will increase. At the moment I am in the process of building such a system for proof of concept. I don’t have anything to show just yet. But I’m getting there.
This past week I was introduced to a gentleman who is working on something similar. However, we have different ideas about what will work. He too sees tags as the way to go but the coding on the tags is where we disagree. I am in favor of a simple binary code such as a bar or QR code where you only need to detect the presences or absence of a strip or block. He, on the other hand, believes that colored bars should be used.
My take on this is that in using the color you’ve gone from having one problem to now having two. Now you’ve got to detect both the presence of the bar or block as well as the color. The advantages I can see with such a system is that your codes could be made smaller. However, changes in lighting and the effect of fading will both cause issue in identifying the tag codes correctly.
Meeting this person, however, resurrected my desire to develop such a system. To that end, I am starting over. I’m going to try and go from identifying photos of cows to ear tags from a long distance and at various angles. Testing various codes with error correction and then see if we can try them out in a herd.
I’d love to hear from ranchers and cattlemen who may have something to add to this project. So feel free to leave comments.