Rail fastener defect recognition algorithm based on deep learning

Abstract

The invention discloses a rail fastener defect recognition algorithm based on deep learning. The rail fastener defect recognition algorithm comprises the following steps that 1, a fastener detection system is adopted; step 2, data preprocessing is carried out; and step 3, feature extraction and classification is carried out. According to the method, the position of the fastener area is verified byutilizing priori knowledge, so that the accuracy of fastener positioning is ensured; the elastic strip sub-images is extracted by using the depth information of the three-dimensional data, and the information of the model input image is successfully simplified; in addition, in order to solve the problem that the number of positive and negative samples is unbalanced, the invention provides a method for creating real simulated fastener fracture data. In the aspect of feature extraction and classification, the feature automatically extracted from the training data by using the deep convolutionalneural network is stronger in expressive force and higher in stability, and the strategy of training feature extraction and classification at the same time is more beneficial to improving the accuracy and stability of the algorithm.

Description

technical field [0001] The invention belongs to the technical field of high-speed railway repair and maintenance, and relates to a track fastener defect identification algorithm based on deep learning. Background technique [0002] In recent years, high-speed railways have developed rapidly in my country. Statistics from China Railway Corporation show that since the development of high-speed railways began in 2008, the total mileage of high-speed railways has reached 25,000 kilometers. Fasteners are one of the important components of high-speed railway track structure, and their long-term and reliable service performance is an important factor for the smooth and safe operation of high-speed trains. However, on-site investigations found that due to factors such as high-frequency train loads and maintenance operations, high-speed railway fasteners often suffer damage such as missing and broken spring bars. The failure of fasteners will affect the dynamic response of the track...

AI Technical Summary

Problems solved by technology

However, two-dimensional image data has inherent defects that cannot be overcome, such as image noise caused by uneven illumination caused by ambient light, quantization noise, etc.

Two-dimensional images mainly provide grayscale and grayscale gradient information. Uneven illumination and noise will seriously affect the feature extraction in the later stage, resulting in low robustness of features.

[0003] Since existing fastener detection algorithms use artificially designed features, such as orientation fields, pyramidal gradient direction histograms, etc., even with the best nonlinear classifiers for feature classification, the accuracy and stability of fastener detection still needs to be further improved. improve

The artificially designed features have the following disadvantages: 1) It is a low-level feature and has insufficient expressive ability for fasteners; 2) Its separability is poor, especially when the data quality is poor, resulting in a high classification error rate

Images