Railway Accident Reduction By Passenger Detection Using Machine Learning Techniques

Railway is the national transporter for many countries with massive rail route networks. The operations and maintenance of such a huge network that is spread over a vast geographical area is an immense task. Currently operations, safety and maintenance-related tasks in railways are done manually or with the help of electronic devices. Automatic Train Protection (ATP) is designed to enhance the safety of a train by applying brakes automatically when it senses any wrong operation or event, but it lacks machine learning techniques like object/obstacle detection algorithms. It is not taking care of the passengers, other obstacles, or other threats coming in between sections. ATP cannot detect operational threats created by Passengers, illegal activities, and Animals. Computer vision can easily detect such threats and Passengers with its advanced algorithms. So it can be the ultimate solution to drawback of ATP. With the use of high computational platforms and machine learning techniques such SSD, YOLO and others passenger detection can be done faster, which can easily increase passengers and operational safety of train. The developed machine learning model will be transferred to a computational system for testing and development. Detected passengers will be then further classified into major and minor threat categories for better control and assistance to driver. A path free from passengers and other obstacles will be predicted and will be used to assist driver of train to increase efficiencies.