Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Track steel surface defect segmentation method based on feature pyramid and neural network

A feature pyramid and convolutional neural network technology, applied in the field of machine vision and deep learning, can solve problems such as long time and inability to use real-time detection, achieve the effect of small number of samples, help network learning, and improve network learning ability

Pending Publication Date: 2022-03-22
SOUTH CHINA UNIV OF TECH +1
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, deep learning requires good hardware conditions to support, and the time for training and testing of large networks is very long, which cannot be used for real-time detection

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Track steel surface defect segmentation method based on feature pyramid and neural network
  • Track steel surface defect segmentation method based on feature pyramid and neural network
  • Track steel surface defect segmentation method based on feature pyramid and neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0083] This embodiment mainly proposes an image segmentation technology combining traditional feature extraction technology and deep learning for the surface of rail steel. First, construct an image pyramid, then extract multi-scale features, and then use multi-scale features as the input of convolutional neural network, train Convolutional neural network. The trained convolutional neural network is able to complete the task of defect segmentation on rail steel surfaces.

[0084] figure 1 It is a flow chart of a method for segmenting surface defects of rail steel based on feature pyramid and convolutional neural network disclosed in this embodiment, and will be described through specific embodiments below. A method for surface defect segmentation of rail steel based on feature pyramid and convolutional neural network, the specific steps are as follows:

[0085] S1. Taking the surface image of rail steel as the input image I 0 , through the input image I 0 Sequentially perf...

Embodiment 2

[0145] In this embodiment, the data set contains a total of 67 samples, 40% or 27 samples are randomly selected as the training set, and the remaining 40 samples are used as the test set, and each sample contains one or more defects. In the examples, three methods based on convolutional neural networks are compared with the method proposed by the present invention, all methods use the same training set and test set. The three methods used for comparison are Unet++, Deeplab, and Segnet respectively, and the method proposed by the present invention is denoted as PFCNN. The measurement results of different methods are shown in Table 1 below:

[0146] Table 1. Evaluation index table of different methods

[0147]

[0148] Where PR, RC, FM are defined as follows:

[0149] PR=TP / (TP+FP); RC=TP / (TP+FN); FM=2×PR×RC / (PR+RC)

[0150] Among them, TP represents the number of defective pixels predicted to be defective, FP represents the number of non-defective pixels predicted to be d...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a track steel surface defect segmentation method based on a feature pyramid and a convolutional neural network. The method comprises the following steps: firstly, constructing a five-layer image pyramid; extracting five feature maps from the image of each layer of the pyramid, and amplifying the feature maps to the original size; and inputting all the feature maps into a lightweight convolutional neural network for training and prediction. The convolutional neural network only comprises 12 convolutional blocks, the parameter quantity is obviously reduced compared with the existing common network, and the training time and the test time are shortened compared with the existing common network; the convolutional neural network adopts a binary cross entropy function and an IOU function as loss functions, the binary cross entropy function can improve the classification capability of a single pixel, and the problem that the number of positive and negative pixels is unbalanced can be solved by designing a relatively high positive sample weight; and the IOU function can improve the accuracy of predicting the shape of the defect.

Description

technical field [0001] The invention relates to the technical fields of machine vision and deep learning, in particular to a method for segmenting surface defects of rail steel based on feature pyramids and convolutional neural networks. Background technique [0002] The development of rail transit is facing the challenge of increasing speed and load, which greatly increases the pressure on rail transit. Long-term operation will cause the surface of the rail to heat up and wear out. Defects such as rail wear and tear are potential safety hazards that need to be resolved urgently. In the past, defects on rails were detected by experienced inspectors. Manual detection not only requires a lot of human resources, but also has disadvantages such as time-consuming and low precision. Therefore, there is a huge market demand for automatic non-destructive testing systems. [0003] Over the past decade, vision-based defect detection methods have been applied to various industrial ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/11G06T7/00G06T5/00G06T3/40G06N3/04G06N3/08G06V10/46
CPCG06T7/11G06T7/0008G06T3/40G06N3/08G06T2207/20081G06T2207/20084G06T2207/30136G06N3/045G06T5/90Y04S10/50
Inventor 刘屿萧华希兰炜圣陈子维
Owner SOUTH CHINA UNIV OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com