Date: 17/06/2026
Hour: 13.30
Student Name & ID: Mehmetali BALABAN & 245105114
Supervisor: Prof.Dr. Ömer KARAL
Topic: Condition Monitoring of Railway Signaling Equipment via Image Processing: A Deep Learning Approach to Signal Lamps
Link or Room: A516
Abstract: The increasing operational density and strict safety requirements in railway transportation present significant opportunities for the application of artificial intelligence in infrastructure condition monitoring processes. Conventional maintenance and diagnostic approaches rely on labor-intensive manual inspections and invasive sensor hardware; these methods are both time-consuming and generate high field costs. In particular, detecting physical deformations of signal lamps, which are critical optical warning devices governed by SIL-4 safety standards, remains a challenging task demanding considerable operational expertise. This study proposes a deep learning-based framework for the automated condition monitoring and fault detection of signal lamps using non-invasive (contactless) image processing techniques. Instead of merely performing binary presence or absence classification, the proposed system operates on digital metrology principles to quantitatively analyze photometric brightness loss in LED modules and pole inclination angles. A convolutional neural network (CNN) architecture is employed to extract discriminative features from visual data, enabling high-accuracy measurements. The proposed approach aims to reduce diagnostic subjectivity, minimize the need for on-site interventions, and serve as a reliable decision support tool for railway operators in the early detection of equipment degradation.