The software design consists of multiple subsystems: the object detection algorithm, the image data input stream, and flight control algorithm. These subsystems work together to help the drone recognize the first aid it and fly towards it.
This model was created using different technologies: OpenCV, YOLOv5, and PyTorch. It detects the first aid kit using the model from a python script. The algorithm is working consistently with test images. When the images come from the drone at more variable distances the results a bit less promising, but the model will be trained further to be more reliable.
The image data stream is being sent through the the DJI app to the flight controller algorithm. After the data stream is sent the object detection algorithm is analyzing if the first aid kit is in each frame.
Currently, the flight control algorithm is not being implemented. A manual operator is flying the drone during the testing phases at the moment. Eventually when the other subsystems are functioning appropriately, the flight control algorithm will be implemented and tested.