Introduction

INTRODUCTION

Railway transportation is one of the most economical and low carbon footprint ways of moving goods and passengers. However, the initial investment required in a railway network is high. Further, railway networks need periodic maintenance for safety and to avoid accidents due to wear and tear of rail tracks which occurs due to passing trains. Freight or goods trains which have high per axle load can lead to faster wear out of the rail tracks. In the US alone, an average of 900+ accidents are reported annually due to track defects.

As the railroad industry rapidly transforms to achieve more loading capacity at higher operating speeds, the traditional ways of track inspection become a bottleneck. The use of Computer Vision based detection techniques along with data science have the potential to detect and predict rail defects and help in delivering a safer transport system.

Rail Track Basics:

The track rails are fixed at a specific distance known as gauge, with the help of sleepers and fasteners as shown in Figure 1. The inner side of the rails is known as Gauge side and the outer side as Field side. The Gauge side of the rail comes in contact with the train wheels. Therefore, Gauge is the side that gets worn out over time.

Table below shows some of the common surface defects on a rail track surface.