An anisotropic heat diffusion model for enhancing the extraction of underground rail track fasteners under extremely low and uneven illumination conditions

Abstract

During the past two decades, subway systems have become one of the most dominant infrastructural developments in China at an unprecedented pace and scale. More than 60 metro lines in 25 cities have been completed, transporting more than 70 million passengers daily. Operating the subway systems safely and efficiently is a continuously pressing demand from both the management companies and the public. Although many automated or semi-automated methods for extracting critical components of the rail track systems, e.g. rail, fastener, sleeper, etc., have significantly improved the productivity of routine inspection, the unique challenges posed by the subway systems have hindered these existing methods from successful implementation because of the extremely low illumination in the underground environment, whereas additional artificial lighting often poses extremely uneven illumination. In this study, a generalized local illumination adaptation model using an anisotropic heat equation is proposed to dynamically adjust the acquired rail track images with extremely low and uneven illumination conditions. An integration flow is then proposed to seamlessly incorporate the proposed model into the state-of-the-art automated fastener detection algorithms. The results show that the proposed local illumination adaptation model can significantly improve the performance of the tested state-of-the-art fastener detection algorithms when they are applied to the images collected in the environment with extremely low and uneven illumination conditions, e.g. subway systems.