Vision-based UAV Localization without GPS

Created using ChatSlide

This project focuses on enhancing UAV localization in GPS-denied environments using a vision-based approach. It introduces a novel system that combines image-map matching through Convolutional Neural Networks (CNNs) with a ResNet50 encoder for robust feature extraction. The integration of a Kalman filter ensures trajectory precision. The system demonstrates superior performance over existing solutions like TransGeo, showing high accuracy in previously untested environments. The project...