A Unified Framework for Fault Detection of Freight Train Images Under Complex Environment

This paper proposes a novel unified framework for fault detection of the freight train images based on convolutional neural network (CNN) under complex environment. Firstly, the multi region proposal networks (MRPN) with a set of prior bounding boxes are introduced to achieve high quality fault proposal generation. And then, we apply a linear non-maximum suppression method to retain the most suitable anchor while removing redundant boxes. Finally, a powerful multi-level region-of-interest (ROI) pooling is proposed for proposal classification and accurate detection. The experimental results indicate that the proposed method can achieve high performance on four typical fault benchmarks, substantially outperforming the state-of-the-art methods.

1. Introduction

Fault detection for the vehicle braking and steering systems is an important routine maintenance task to ensure the security of freight trains. For a long time, it has been performed by skilled workers, which has many drawbacks such as low detection probability and poor efficiency. For the efficiency and convenience of detection, vision -based fault detection of freight trains is getting more and more attention [1], , [3]. The vehicle braking and steering systems contain many important parts such as cut-out cocks, dust collectors, bogie block keys, and fastening bolts [4], [5], etc. Their technical status, performance and stability are directly related to the train operation safety.