A conceptual model to detect seat damages in train compartments

Abstract

Railway transport is one of the most popular public transportation mechanisms, which serves about seven million passengers daily. The Department of Railways has provided various provisions and safety regulations to provide its passengers with a superior and convenient service. However, some irresponsible passengers are not following the rules and regulations imposed by the department and damaging the various parts of the train compartments. Unscrupulous damage and writing junk on the seat are significant issues. There is no proper mechanism to identify the passenger who damages the train seat. Currently, the Department of Railways tries to control these problems by finding and controlling them by using their employees as inspectors, but it is not up to their expectations. This study aimed to address the problems related to seat damages by using the images retrieved from train compartment cameras, which are permanently fixed inside train carriages. The system will identify the damage done by a passenger to the train seat by comparing the seat images at regular time intervals using image processing techniques and Convolutional Neural Networks. When there are damages, the department staff will be notified with the proof of damages and passenger information. A pilot project is under development for long-distance first-class train compartments. The proposed model has currently achieved over 94% training accuracy while identifying damages to the seats. The proposed model will be further enhanced to identify all kinds of damages to the train compartments across all classes in the train. The improved model will help the railway department protect its train compartments while providing an enhanced experience to its passengers.