The autonomous monitoring system consists of Avular’s Vertex drone platform equipped with a camera and artificial intelligence from ADI.
To facilitate navigation inside the greenhouse, Avular will install beacons of the Spectrum Ultra-Wide Band navigation system in the greenhouses, which allows the drone to fly autonomously through the greenhouse and to collect image data of the plants. These images will be analyzed by the artificial intelligence from ADI to identify crop growth stages and plant diseases.
Flying inside greenhouses is challenging. First of all, the drone should be very good at recognizing glass walls and roofs to avoid crashes; something that is very difficult with visual sensor types. To solve this problem, Avular will test different sensors to optimize this on the drone. Another challenge is the warm and humid climate in greenhouses. The drone and the navigation infrastructure need to be adapted for long term deployment under these conditions. This requires redesign of the mechanics and electronics of these systems.
When the drone is ready to fly in the greenhouse, it will need something to stabilize the camera during the flight. Otherwise, the camera may record unsharp images when the drone makes turns or other sudden movements. To prevent this from happening, Avular will develop a camera module with gimbal to facilitate high quality data collection with the drone.
In turn, ADI will develop the artificial intelligence to analyze the visual data acquired by the drone. To do so, the AI will be trained to recognize healthy and unhealthy plants. This is done by feeding the AI a large quantity of training pictures of healthy plants and plants that are infected with the top-5 plant diseases.
So far, Avular has worked on height estimation of the drone in the greenhouse. Accurate height estimation is vital to prevent the drone from flying through the roof of the greenhouse. We have discovered that a barometer can’t be used for this. We now integrated a stereovision camera on the drone for height estimation. This seems to work very well and will soon be tested in the greenhouse.
Meanwhile, ADI has been working on an elaborate simulation to calculate the effects of greenhouse conditions on the UWB system that is used for navigation. The results of this simulation will be used to improve the UWB system.
Hopefully, the autonomous monitoring system will be finished by the end of 2020. In the future, the functionalities of this autonomous monitoring system can be extended by gathering more types of data or by equipping the drone with end effectors that allow precision farming.
Project page at ADI
Partner: ADI applied drone innovations