An Intelligent Identification Method of Railway Subgrade Diseases Based on Radar Data

Abstract

The invention discloses a method for intelligent identification of railway roadbed diseases based on radar data. Convert the data and label files, divide the training set and test set, expand and normalize the training set and send it to the convolutional neural network, output the disease type, position coordinates and confidence, and obtain the railway roadbed disease detection through iterative calculation of the gradient descent method model, and use the mean average precision and the number of frames per second as evaluation indicators, and finally convert the coordinates of the disease position into the starting and ending mileage and the starting and ending depth of the disease. Compared with the existing image-based railway roadbed disease detection method, this method makes full use of the original radar data and combines it with the convolutional neural network, and provides the mileage and the start and end of the disease while intelligently identifying the railway roadbed disease Depth, to meet the engineering needs.

Description

technical field [0001] The invention relates to the technical fields of railway embankment disease detection and radar signal intelligent identification, in particular to an intelligent identification method for railway embankment diseases based on radar data. Background technique [0002] With the development of my country's railway technology, railway subgrade disease has become a hidden danger affecting the safe operation of railways, which is critical to driving safety. The vehicle-mounted geological radar detection method has been paid attention to in the detection of railway subgrades in various countries due to its advantages of lightness, quickness, strong anti-interference ability, high resolution, fast and non-destructive. [0003] At present, the identification of railway embankment diseases mainly relies on the experience of engineers and technicians to manually judge GPR images, which is costly, inefficient, and the standard is not fixed. Existing intelligent r...

AI Technical Summary

Problems solved by technology

[0003] At present, the identification of railway embankment diseases mainly relies on the experience of engineers and technicians to manually judge ground-penetrating radar images, which is costly, inefficient, and the standard is not fixed.

Existing intelligent recognition technologies are mostly based on radar images, such as the combination of traditional machine learning methods such as support vector machines and shallow neural networks, and convolutional neural network recognition technology based on candidate areas. The specific location in the railway cannot meet the needs of detecting the mileage and depth of the disease

Images