Integration of Drone and Computer Vision for Object Detection on Railway Tracks

This paper explores the feasibility of drone flights integrated with computer vision technology for object detection on railway tracks and, thus, aims overall to enhance the workflow required for monitoring tasks. The conceptual model relies on employing existing technology in tracking railway lines for semi-autonomous navigation and artificial intelligence for visual identification. This was demonstrated by a stepwise operation of a drone flight via a navigation algorithm and visual data processing procedure via a machine learning model trained by a manually created supervision library. Experimentation flights showed navigation following railway tracks, executed at 70% for routes of short distances of 100 m or less, and object detection performed at various levels of confidence. The study carried out here highlights the feasibility of such integration and provides a basis for practical implementation.