HORIZON DETECTION using Machine Learning

Abstract

This project focuses on utilizing machine learning techniques for horizon detection, with applications ranging from planetary rover localization to flight control and port security. Traditional methods like GPS are unavailable in certain scenarios, making horizon detection a critical visual cue for precise position and orientation estimation. The proposed approach involves Canny edge detection, followed by the classification of edges as either "horizon" or "non-horizon" using a Support Vector Machine (SVM) trained on ground truth data and SIFT (Scale-Invariant Feature Transform) features. When applied to new images, this process identifies a consistent horizon line, offering a robust and adaptable solution for addressing \complex real-world challenges.