WILDLABS.NET Virtual Meetup: Drones
I haven't posted anything on drones since the August event at the Olympic Park, mainly due to other work getting in the way. However, the WILDLABS.NET organisation had a virtual meeting about the use of drones in wildlife conservation last week. The details are on their website here: https://www.wildlabs.net/community/thread/809
They had three short presentations about the ArduPilot software, using infra-red cameras to detect wildlife and finally the work being done by Conservation International. This was a bit of a reminder that I need to progress some of the drone projects that I've been working on, as the infra-red detection is something that I've been interested in for a number of years. They showed an example of offline annotated infra-red video with the animal species localised in a bounding box and classified according to type. If you were very quick, you could see them using YOLO, which is a carbon copy of the system that we used when we tackled the Lockheed Martin, Alpha Pilot AI drone racing challenge this time last year. However, while they're using large drones and high resolution cameras, we've been slowly developing the micro drone technology on the basis that they have less of an impact on wildlife and you can get them into places that their bigger brothers can't go. Also, with the new DRES regulations becoming law at the end of this month, our drones are licence exempt as they fall below the minimum weight threshold. Actually, there was a lot of discussion around the legal restrictions of flying drones, both in this country and abroad. I've been flying radio controlled for over thirty years, so I've read and digested CAP722, and already have my operator ID and flyer ID, even though I'm exempt until the end of the year.
The result of watching this discussion was to make me realise how much great technology I have sitting on the desk in front of me right now. There is a 6g LIDAR, a similarly sized micro infra-red camera that I'm working on, the BBC micro-bit hardware that I need to finish the flight controller for, plus half a dozen micro drones and a Taranis controller. Most interestingly, though, there is the kit for the AI drone racing project which still hasn't been fully tested yet. I have a Movidius neural network on a stick, which can run the drone racing gate detection and localisation algorithm that we developed for the Alpha Pilot project. Now, while AI drone racing might sound like fun (OK, it is fun), there is a serious side too. This type of autonomy can be used as a safety feature to prevent the drone from crashing into anything it shouldn't (i.e. the protected species of bird that you're trying to save), while also being able to navigate and map in real-time, or perform search and rescue. There's at least one example of a drone swarm being let loose in a forest to find someone who is lost and then there's the tunnel mapping example from the Prometheus movie.
So, this post has been about what other people are doing, but it's provided me with just the inspiration that I needed to get going again after a long break. The future is all about us progressing our micro drone expertise, and I've got a few ideas.
They had three short presentations about the ArduPilot software, using infra-red cameras to detect wildlife and finally the work being done by Conservation International. This was a bit of a reminder that I need to progress some of the drone projects that I've been working on, as the infra-red detection is something that I've been interested in for a number of years. They showed an example of offline annotated infra-red video with the animal species localised in a bounding box and classified according to type. If you were very quick, you could see them using YOLO, which is a carbon copy of the system that we used when we tackled the Lockheed Martin, Alpha Pilot AI drone racing challenge this time last year. However, while they're using large drones and high resolution cameras, we've been slowly developing the micro drone technology on the basis that they have less of an impact on wildlife and you can get them into places that their bigger brothers can't go. Also, with the new DRES regulations becoming law at the end of this month, our drones are licence exempt as they fall below the minimum weight threshold. Actually, there was a lot of discussion around the legal restrictions of flying drones, both in this country and abroad. I've been flying radio controlled for over thirty years, so I've read and digested CAP722, and already have my operator ID and flyer ID, even though I'm exempt until the end of the year.
The result of watching this discussion was to make me realise how much great technology I have sitting on the desk in front of me right now. There is a 6g LIDAR, a similarly sized micro infra-red camera that I'm working on, the BBC micro-bit hardware that I need to finish the flight controller for, plus half a dozen micro drones and a Taranis controller. Most interestingly, though, there is the kit for the AI drone racing project which still hasn't been fully tested yet. I have a Movidius neural network on a stick, which can run the drone racing gate detection and localisation algorithm that we developed for the Alpha Pilot project. Now, while AI drone racing might sound like fun (OK, it is fun), there is a serious side too. This type of autonomy can be used as a safety feature to prevent the drone from crashing into anything it shouldn't (i.e. the protected species of bird that you're trying to save), while also being able to navigate and map in real-time, or perform search and rescue. There's at least one example of a drone swarm being let loose in a forest to find someone who is lost and then there's the tunnel mapping example from the Prometheus movie.
So, this post has been about what other people are doing, but it's provided me with just the inspiration that I needed to get going again after a long break. The future is all about us progressing our micro drone expertise, and I've got a few ideas.
Comments
Post a Comment