Steel rail surface defect identification and classification method

Abstract

The invention provides a steel rail surface defect identification and classification method, which is based on unmanned aerial vehicle monitoring image and depth forest model analysis, and comprises the following steps of: firstly, preprocessing an unmanned aerial vehicle monitoring image; secondly, extracting steel rail surface defects based on a proportional reinforcement maximum entropy threshold algorithm; and finally, realizing small sample defect data classification by utilizing a depth forest method. The steel rail surface defect identification and classification method provided by theinvention has the advantages and positive effects that the blind zone problem of a traditional detection mode in the field of steel rail surface defect detection is solved, the improved method for processing the unmanned aerial vehicle image is provided, and a certain promotion effect is achieved on track defect detection by utilizing the unmanned aerial vehicle; and the defect classification method suitable for the small samples is provided, and the problem of insufficient sample data volume at present can be solved on the basis of ensuring the accuracy.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for identifying and classifying rail surface defects. Background technique [0002] The existence of track surface defects leads to unstable operation of the train, which greatly damages the wheels, wheel sets and bogie components, which not only shortens the service life of each part of the train, but may even threaten the safe driving of the train, such as causing railway accidents such as train derailment . Therefore, effective detection of track surface defects is an essential measure to ensure the safe and reliable operation of the railway system. [0003] At present, there are not many methods for rail surface defect detection at home and abroad. The methods commonly used in various countries are mainly concentrated in the modern methods of visual inspection, ultrasonic testing, and eddy current method. With the rapid development of automation and digita...

AI Technical Summary

Problems solved by technology

First of all, the deep neural network needs a large number of samples to train. For small samples, the results obtained by using the deep learning network for training are not ideal.

Second, the hyperparameters of the deep learning network are many and complex, and the learning process depends on the adjustment of hyperparameters

For rail surface defect images, due to the fact that there are relatively few surface defects in the track lines currently in normal operation, and due to the limited range of data collection, it is impossible to obtain enough images for deep learning network training. sample

Images