Train Track Crack Prediction Using CNN with LeNet - 5 Architecture

Train track crack detection is a process of identifying cracks in the structure of railway tracks. Railways are major modes of transport in India. The tracks must be in good condition for trains to have safe voyages. Cracks that appear on the tracks are often due to heat and other natural causes. At present these cracks are identified manually by railway personnel by inspecting them at regular intervals. This process is not effective as it consumes more time and there is an increased chance of leaving the cracked track undiscovered. The aim of this research work is to avoid the derailment of trains and reduce the cost and time that happens due to the cracks. This work proposed a technique for recognizing railway track cracks by combining Convolutional Neural Networks with image pre-processing techniques. Observations indicate that neural networks are capable of capturing the colours and textures of lesions related to respective railway track breaks with 94.6% accuracy.