ETD

• Annotation tool
• CNN
• Data Augmentation
• Deep Learning
• Hough Transform
• MATALB App Designer
• MATLAB ImageLabeler
• Preprocessing
• S-curves
• Satellite Images
• YOLO

Riassunto tesi:
This thesis work aim to put basis towards the developement of a neural network based on the Hough Transform that can visually detect s-curves in radiofrequency satellite images. In this work it has been developed an annotation tool in MATLAB App Designer in order to be able to create a ground truth for the Hough Transform. It has been also used another MATLAB tool to manually annotate images in order to train a state-of-the-art object detection model, like YOLOv11, to use as bechmark for the future developed neural networks.
Preprocessing and data augmentation techniques has been applied to the dataset of simulated images in order to get around the problem of lacking of real images.There has been used data augmentation techinques that aims to make simulated images closer in their appearences to real images. Models obtained are then tested on the real images to see the if data augmentation techninques were usefull to improve detection capabilities on the real use cases images.