Intelligent Image-Based Railway Inspection System Using Deep Learning-Based Object Detection and Weber Contrast-Based Image Comparison

Jang, Jinbeum; Shin, Minwoo; Lim, Sohee; Park, Jonggook; Kim, Joungyeon; Paik, Joonki

doi:10.3390/s19214738

Cited by 25 publications

(15 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The detection error of ultrasound techniques is large with certain crack angles or defect sizes smaller than 5% of the railhead area [ 13 ]. In contrast, some non-contact technologies like cameras can continuously capture the pictures that record the locations and sizes of the defects [ 14 ]. The significant issue is that complicated background noises pollute the defect images.…”

Section: Introductionmentioning

confidence: 99%

Wavelet Scattering and Neural Networks for Railhead Defect Identification

Yang

2021

Materials

View full text Add to dashboard Cite

Accurate and automatic railhead inspection is crucial for the operational safety of railway systems. Deep learning on visual images is effective in the automatic detection of railhead defects, but either intensive data requirements or ignoring defect sizes reduce its applicability. This paper developed a machine learning framework based on wavelet scattering networks (WSNs) and neural networks (NNs) for identifying railhead defects. WSNs are functionally equivalent to deep convolutional neural networks while containing no parameters, thus suitable for non-intensive datasets. NNs can restore location and size information. The publicly available rail surface discrete defects (RSDD) datasets were analyzed, including 67 Type-I railhead images acquired from express tracks and 128 Type-II images captured from ordinary/heavy haul tracks. The ultimate validation accuracy reached 99.80% and 99.44%, respectively. WSNs can extract implicit signal features, and the support vector machine classifier can improve the learning accuracy of NNs by over 6%. Three criteria, namely the precision, recall, and F-measure, were calculated for comparison with the literature. At the pixel level, the developed approach achieved three criteria of around 90%, outperforming former methods. At the defect level, the recall rates reached 100%, indicating all labeled defects were identified. The precision rates were around 75%, affected by the insignificant misidentified speckles (smaller than 20 pixels). Nonetheless, the developed learning framework was effective in identifying railhead defects.

show abstract

Section: Introductionmentioning

confidence: 99%

Wavelet Scattering and Neural Networks for Railhead Defect Identification

Yang

2021

Materials

View full text Add to dashboard Cite

show abstract

“…The semantics of the highlighted information is recognized by artificial intelligence. In [6], to realize the intelligent appearance inspection of railway infrastructure, the author proposed a new railway inspection system based on deep convolutional neural network and computer vision-based image comparison method to accurately find the facility and detect its potential defect. In [7], the author proposed a real-time automatic integrated circuit mark inspection system based on an embedded platform to identify the rotation and position of the IC chip.…”

Section: Introductionmentioning

confidence: 99%

Intelligent Image Detection System Based on Internet of Things and Cloud Computing

admin¹,

Algrnaodi²

2021

JISIoT

View full text Add to dashboard Cite

Images are the most intuitive way for humans to perceive and obtain information, and they are one of the most important sources of information. With the development of information technology, the use of digital image processing methods to locate and identify targets is widely used, so it is particularly important to detect the targets of interest quickly and accurately in the image. The traditional image detection system has the problems of low detection accuracy, long time consumption, and poor stability. Therefore, this paper proposes the design and research of artificial intelligence image detection system based on Internet of Things and cloud computing. The system designed in this article mainly includes three links, namely: image processing analysis design link in cloud computing environment, image feature collection module design link, and image integration detection link. The main technologies used in image processing and analysis in the cloud computing environment are virtualization technology, distributed massive data storage, and distributed computing. In the image feature collection module, before the image is input to the neural network, it is necessary to perform preprocessing operations on the distorted image and perform perspective correction; then use the deep residual network in deep learning to extract features. Finally, there is the image integration detection link. First, the target category judgment and position correction are performed on the regions generated by the candidate region generation network, and then the integrated image detection is performed through the improved target detection method based on the frame difference method. Through simulation experiments, compared with the traditional image detection system, the speed advantage of the artificial intelligence image detection system designed in this paper is obvious in the case of a large increase in the number of images. On images at different pixel levels, the accuracy of the image detection system proposed in this paper is always higher than that of traditional image detection systems, and the CPU usage and memory usage are at a lower level. In addition, within three months, the stability is also at a relatively high level of 0.9.

show abstract

“…Min et al [ 21 ] and Wei et al [ 22 ] proposed a rail surface detection system based on machine vision. Jang et al [ 23 ] Shang et al [ 24 ] Faghih-Roohi et al [ 25 ] and Song et al [ 26 ] detected the rail surface damage by using the deep learning methods. In recent years, measurement methods based on 3D laser vison have become widely used for rail surface detection.…”

Section: Introductionmentioning

confidence: 99%

Establishment of the Complete Closed Mesh Model of Rail-Surface Scratch Data for Online Repair

Guo

Huang

Liu

et al. 2020

Sensors

View full text Add to dashboard Cite

Rail surface scratching occurs with increasing frequency, seriously threatening the safety of vehicles and humans. Online repair of rail-surface scratches on damaged rails with scratch depths >1 mm is of increased importance, because direct rail-replacement has the disadvantages of long operation time, high manpower and high material costs. Advanced online repair of rail-surface scratch using three-dimensional (3D) metal printing technology such as laser cladding has become an increasing trend, desperately demanding a solution for the fast and precise establishment of a complete closed mesh model of rail-surface scratch data. However, there have only been limited studies on the topic so far. In this paper, the complete closed mesh model is well established based on a novel triangulation algorithm relying on the topological features of the point-cloud model (PCM) of scratch-data, which is obtained by implementing a scratch-data-computation process following a rail-geometric-feature-fused algorithm of random sample consensus (RANSAC) performed on the full rail-surface PCM constructed by 3D laser vision. The proposed method is universal for all types of normal-speed rails in China. Experimental results show that the proposed method can accurately acquire the complete closed mesh models of scratch data of one meter of 50 Kg/m-rails within 1 min.

show abstract

Intelligent Image-Based Railway Inspection System Using Deep Learning-Based Object Detection and Weber Contrast-Based Image Comparison

Cited by 25 publications

References 22 publications

Wavelet Scattering and Neural Networks for Railhead Defect Identification

Wavelet Scattering and Neural Networks for Railhead Defect Identification

Intelligent Image Detection System Based on Internet of Things and Cloud Computing

Establishment of the Complete Closed Mesh Model of Rail-Surface Scratch Data for Online Repair

Contact Info

Product

Resources

About