Research On High-speed Railway Fasteners Crack Detection Algorithm

The rail fastener is an important facility that directly interacts with the high-speed train,the fatigue fracture and failure of the rail fastener seriously affect the safe operation of the high-speed train.At present,the manual detection method is not only inefficient,but also has a high detection error rate.It can be seen that it is of great significance to study the fast detection algorithm for high-speed rail fasteners for the safe operation of high-speed railways.In this paper,based on the characteristics of cracks on fasteners,a method for detecting cracks on rail fasteners based on phase congruency is proposed.Firstly,the Hough Circle transform is used to locate the circular nut above the rail fastener,and then realizes the precise positioning of the crack region according to the invariance of relative position of the circular nut and the crack region.Secondly,the phase congruency and threshold segmentation are performed on the image of crack region,and a series of morphological processing is used.Finally,the binary image is projected vertically,the feature matrix obtained by the projection is input into the support vector machine classifier for training and testing,and the crack detection of the fastener is realized.In this paper,1500 real images collected by the actual track detection vehicle are used for experiments.The accuracy of the crack detection method based on phase congruency can reach 87.47%.Compared with the traditional image edge detection algorithms Sobel and Canny,it shows stronger robustness and higher detection rate.Since the convolutional neural network has achieved remarkable results in various fields such as image processing and speech recognition in recent years,this paper proposes another method based on convolutional neural networks to detect cracks in high-speed railway fasteners,after the feature are is located,the convolutional neural network is used to detect cracks on fasteners.Training convolutional neural networks requires a large amount of data,we use a sliding window detection scheme to train convolutional neural networks with a large number of small blocks extracted from the original image which greatly improving the stability and accuracy of convolutional neural networks.In addition,this paper proposes an optimization method for the insufficiency of convolutional neural network detection results.Using the SVM classifier based on HOG feature to optimize the detection results of convolutional neural network,which greatly enhances the robustness of the detection algorithm.In the experiment,the effectiveness of the proposed method is verified by using 1500 image collected by the actual track detection vehicle.The improved CNN detection method can achieve the accuracy of 91.73%.Therefore,the proposed method is more suitable for the detection of cracks in high-speed railway fasteners than traditional detection algorithms.