As railroads worldwide are investing for improving their capacity, efficiency and safety, there is a realization that there is a dearth of a seamless intelligent uniform infrastructure that serves as information management platform for its diverse application needs. One of the primary objectives our project is to explore an integration framework that seamlessly marries the requirements of high-level operation management objectives with idiosyncratic interfaces of multiple low-level inhomogeneous subsystems comprising of sensors, measurements, analytics, communication, and control. The sheer magnitude of the problem and its mission critical nature demands multiple innovations involving capturing disparate data sources and managing multiple analytics for arriving intelligent inferences thereof, and for disseminating the information to the users. Taking the concrete contemporary context of railroad safety requirements and use case scenarios, the IBM SAFEST first-of-a-kind (FOAK) project proposes to develop an automatic problem detection framework for incident prevention and mitigation by leveraging many proven IBM Research technologies including advanced machine learning, data mining, and low latency infrastructure. We believe this framework will serve as a proof point for the interconnected, instrumented and intelligent smarter railroads.
During this project we intend to validate this objective in the concrete context of a specific railroad safety application, namely, track inspection, within travel and transportation industry. Specifically, we aim to detect abnormal spiking and anchor patterns, as well as identifying missing, loose, and broken joint bolts on different classes of tracks, by automatically analyzing videos captured by multiple cameras (up to 6) mounted on a hi-rail vehicle.
The following Research innovations will be leveraged by the project:
Members: Ying Li, Norm Haas, Charles Otto, Yuichi Fujiki, Sharath Pankanti