IOT-Enhanced Computer Vision Framework for Real-Time Railway Track Defect Detection and Classification

Abstract This paper presents an IoT-enhanced computer vision framework that addresses the critical challenge of railway track defect detection and classification in real-time. Railway infrastructure safety remains paramount for passenger security and operational efficiency, with traditional inspection methods proving costly and time-inefficient. Our solution integrates IoT sensor networks with advanced computer vision algorithms to enable continuous monitoring. The framework employs a three-tier architecture: (1) distributed IoT sensors capturing high-resolution visual data, (2) edge computing nodes performing preliminary analysis, and (3) cloud-based deep learning models for comprehensive defect classification. We implemented a modified ResNet-50 CNN architecture trained on a custom dataset of 12,500 railway track images with various defect types. Experimental validation demonstrates remarkable performance with 97.8% accuracy in defect detection and 94.3% precision in classification across seven defect categories. The system operates with a latency of 0.38 seconds, enabling truly real-time monitoring of track conditions. This framework significantly reduces inspection costs while enhancing safety through continuous monitoring, representing a viable solution for next-generation railway maintenance systems. Keywords: Railway track defects, IoT sensors, computer vision, real-time monitoring, deep learning