Railway Freight Wheel Tread Detection Based On Motion Blur Infrared Image

The driving safety detection of railway wagon means a lot to ensure the steadily running of railway wagon.Especially the safety condition of wheelset is a key concern in the process of railway wagon operation and maintenance.At present,wheelset tread damage detection is carried out under non-operating condition,which is slow,inefficient and consumes a lot of manpower and material resources.Therefore,it is still necessary to explore a more rapid,efficient and convenient wheelset tread detection method for railway wagon.Theoretical and experimental research is carried out on the damage mechanism of railway wagon wheelset tread,finite element simulation of temperature field,motion-blurred image restoration,image feature extraction and so on in this paper.In order to detect the damage of wheelset tread in motion state,a blind restoration algorithm based on multi-scale sensing region detection is proposed for the motion-blurred infrared image,and an infrared dim target detection algorithm based on sliding window is proposed for the wheelset tread region.The specific research contents are as follows:(1)Firstly,the causes and characteristics of typical tread damage are analyzed.Then,the tread spalling/shelling process of brake shoe tread braking is studied,and the nonlinear strong coupling characteristics of thermal,mechanical and microstructure with temperature,load and material are analyzed.Then the abnormal distribution of tread temperature is found.The feasibility of detecting tread damage by temperature field variation is demonstrated.Finally,the data acquisition scheme is designed and the characteristics of infrared image are analyzed.(2)The characteristics of transient temperature field distribution of railway wagon wheels braking through brake shoe tread under specified working conditions are studied.Firstly,according to the actual investigation and heat transfer theory,wheel models with or without tread stripping are established,and the boundary conditions and simulation conditions are determined.Then,the temperature field of the wheel is simulated.The temperature variation laws of the wheel in the braking and cooling stages are determined,and the distribution characteristics of the temperature field of the two wheel models are analyzed and compared.Finally,the simulation results show that the temperature distribution of wheel with tread spalling/shelling is different from that without tread spalling/shelling under the same working condition.The maximum temperature at which the tread peeled off reaches 35.3°C at 28 s,while the temperature at the tread surface is 23.3°C,with the difference of 12.0°C.(3)An algorithm of motion-blurred infrared image restoration is proposed to reduce the influence of noise and motion blur on infrared image.Firstly,the image quality reduction model and the image blind restoration algorithm are analyzed.Then,based on the non-uniform blur hypothesis,the blind restoration algorithm based on total variational regularization is optimized,that is,the blind restoration algorithm based on multi-scale sensing region detection.In this algorithm,a multi-scale sensing region detection algorithm is used to judge the blur region in the image,and a two-dimensional maximum entropy image segmentation algorithm is used to extract the blur region,and then a blind restoration method based on total variational regularization is used to restore the blur region.Finally,subjective evaluation method and objective evaluation methods are used to compare the processing results of the two algorithms from the perspectives of noise,image edge and motion blur.The results show that the optimal algorithm has advantages in the blind restoration of motion-blurred infrared images.(4)An algorithm for feature extraction of railway wagon wheel tread is proposed,and the area of tread shelling/spalling is calculated.Firstly,the basic principle of image segmentation method is analyzed.Firstly,the basic principle of image segmentation method is analyzed.Experiments show that the existing method is difficult to detect the feature of tread shelling/spalling due to the improper selection of parameters.Therefore,a new algorithm of infrared dim and small target detection based on sliding window is proposed.Firstly,the basic principle of image segmentation method is analyzed.Experiments show that the existing method is difficult to detect the feature of tread shelling/spalling due to the improper selection of parameters.Therefore,a new method of infrared dim and small target detection based on sliding window is proposed.The method divides the sub-blocks according to certain rules,automatically determines the segmentation threshold based on hypothesis testing,determines the reference point according to the decision conditions and threshold values,and obtains the outer rectangle of the region of interest.Then,the area characteristics of the tread shelling/spalling are determined.Compared with the visible image,the causes of the errors are analyzed,and the accuracy of the extracted features is evaluated.This paper studies the wheel tread of railway wagon in motion,demonstrates the characteristics and differences of temperature field distribution with or without tread shelling/spalling;proposes a restoration algorithm for motion-blurred infrared images to reduce the influence of noise and motion blur in infrared images;proposes feature extraction algorithm of the wheel tread,and calculates the area of the tread shelling/spalling.The dynamic detection on the tread damage of the railway wagon wheel is completed.