Some scripts done for the development of drone automation, some developt at Politecnico di Torino's DRAFT student group.
These scripts were use to develop drone's autonomy using machine learning and computer vision techniques in the ROS/Gazebo simulation Hector Quadrotor.
This folder includes functions used to build a virtual dataset of a quadrotor reaching a gate for drone racing.
To build the dataset, the controller used gets the position of the center of the gate with the informations provided by gazebo and follows the gate's position with a simple PID controller.
The drone's relative position with the gate and the images are then saved.
For the current corner detection script it is supposed that there is only one gate per image and that the gate's color is known.
The script first applies a color segmentation, then a Harris corner detection and then, to get only the 4 gate's corner, a KMeans segmentation.