Research on Rapid Position of Axle Box in High-Noise Infrared Images Acquired by Trackside

Axle box bearing is an important part of the rotating part of the bogie, and its state is related to the running safety of the train. For fault diagnosis and condition monitoring, vibration and temperature are measured from the bogie. However, installing additional sensing devices on the bogie increases manufacturing cost and can only obtain the information of a single point. In this article, we use the infrared monitoring system beside the railway to obtain global temperature information of the axle box. Aiming at a large number of infrared images with high noise caused by motion, this article presents a rapid position method of axle box based on image processing and infrared image characteristics. This article extracted G component as the input image, proposed adaptive temperature multicenter of gravity positioning method and curve detection method with the fitting radius of the boundary as shape descriptor to label the target image, and then used continuous-frame information to rapidly position the axle box based on the target image. The verification is carried out on six groups of real data. The results demonstrate the proposed method has 100% positioning accuracy, and the average time of each group is 11.77 s. The proposed method is more effective for images with high noise and blurred edges caused by motion. And on the premise of good robustness to ambient temperature and slight image deformation, it can also meet the real-time requirements, which can better adapt to the complex environment in practical engineering applications.