Autonomous Robotic Inspection in Tunnels

Protopapadakis, Eftychios; Stentoumis, Christos; Doulamis, Nikolaos; Doulamis, Anastasios; Loupos, Konstantinos; Makantasis, Konstantinos; Kopsiaftis, George; Amditis, Angelos

doi:10.5194/isprs-annals-iii-5-167-2016

Cited by 22 publications

(12 citation statements)

References 24 publications

(10 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Protopapadakis et al 31 introduced an integrated automatic platform for road and railway tunnel inspection. Based on this platform, a three-layer convolutional neural network (CNN) taking 9 × 9 patches as input was used as feature extract.…”

Section: Related Workmentioning

confidence: 99%

“…They showed that CNN is good at detecting thin cracks under varying lighting conditions and is more robust to noise than any traditional methods. As examples of the early deep-learning-based methods, Protopapadakis et al 31 and Cha et al 13 generated a single label for a patch, and sliding window techniques were applied to scan through the whole image. This kind of techniques could only provide information about whether a patch include cracks or not.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Multi-level feature fusion in densely connected deep-learning architecture and depth-first search for crack segmentation on images collected with smartphones

Mei

Gül

2020

Structural Health Monitoring

View full text Add to dashboard Cite

Cracks are important signs of degradation in existing infrastructure systems. Automatic crack detection and segmentation plays a key role in developing smart infrastructure systems. However, this field has been challenging over the last decades due to irregular shape of the cracks and complex illumination conditions. This article proposes a novel deep-learning architecture for crack segmentation at pixel-level. In this architecture, one convolutional layer is densely connected to multiple other layers in a feed-forward fashion. Max pooling layers are used to reduce the dimensions of the features, and transposed convolution layers are used for multi-level feature fusion. A depth-first search–based algorithm is applied as post-processing tool to remove isolated pixels and improve the accuracy. The method is tested on two previously published data sets. It can reach 92.02%, 91.13%, and 91.58% for the first data set, and 92.17%, 91.61%, and 91.89% for the second data set for precision, recall, and F1 score, respectively. The performance of the proposed method outperforms other state-of-the-art methods. At the end of the article, the influence of feature fusion methods and transfer learning are also discussed.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Multi-level feature fusion in densely connected deep-learning architecture and depth-first search for crack segmentation on images collected with smartphones

Mei

Gül

2020

Structural Health Monitoring

View full text Add to dashboard Cite

show abstract

“…Following this success, CNN has revolutionised the field of computer vision and become a state of the art for recognition and detection purpose [5]. In addition, it has been utilised in a wide range of application such as image classification for a large number of classes [9-11], traffic sign recognition [12], medical object classification [13][14][15], face recognition [16,17], and damage detection in structures [18][19][20][21]. In addition to automatic feature extraction, CNN can produce excellent performance for complex image recognition task due to the capability of CNN in exploiting the local spatial correlation between pixels in the image [5].…”

Section: Bridge Load Testingmentioning

confidence: 99%

Deep learning for bridge load capacity estimation in post-disaster and -conflict zones

et al. 2019

View full text Add to dashboard Cite

Many post-disaster and post-conflict regions do not have sufficient data on their transportation infrastructure assets, hindering both mobility and reconstruction. In particular, as the number of ageing and deteriorating bridges increases, it is necessary to quantify their load characteristics in order to inform maintenance and asset databases. The load carrying capacity and the design load are considered as the main aspects of any civil structures. Human examination can be costly and slow when expertise is lacking in challenging scenarios. In this paper, we propose to employ deep learning as a method to estimate the load carrying capacity from crowdsourced images. A convolutional neural network architecture is trained on data from over 6000 bridges, which will benefit future research and applications. We observe significant variations in the dataset (e.g. class interval, image completion, image colour) and quantify their impact on the prediction accuracy, precision, recall and F1 score. Finally, practical optimization is performed by converting multiclass classification into binary classification to achieve a promising field use performance.

show abstract

“…In recent years, relevant scholars have researched the application of deep learning in the detection of cracks at tunnel linings. References [ 25 , 26 , 27 ] achieved good results in terms of accuracy and efficiency. Nevertheless, deep learning algorithms require a large number of labeled images of training, and these images are not easy to obtain.…”

Section: Introductionmentioning

confidence: 99%

Dynamic Partition Gaussian Crack Detection Algorithm Based on Projection Curve Distribution

Xue

Yuan

2020

Sensors

View full text Add to dashboard Cite

When detecting the cracks in the tunnel lining image, due to uneven illumination, there are generally differences in brightness and contrast between the cracked pixels and the surrounding background pixels as well as differences in the widths of the cracked pixels, which bring difficulty in detecting and extracting cracks. Therefore, this paper proposes a dynamic partitioned Gaussian crack detection algorithm based on the projection curve distribution. First, according to the distribution of the image projection curve, the background pixels are dynamically partitioned. Second, a new dynamic partitioned Gaussian (DPG) model was established, and the set rules of partition boundary conditions, partition number, and partition corresponding threshold were defined. Then, the threshold and multi-scale Gaussian factors corresponding to different crack widths were substituted into the Gaussian model to detect cracks. Finally, crack morphology and the breakpoint connection algorithm were combined to complete the crack extraction. The algorithm was tested on the lining gallery captured on the site of the Tang-Ling-Shan Tunnel in Liaoning Province, China. The optimal parameters in the algorithm were estimated through the Recall, Precision, and Time curves. From two aspects of qualitative and quantitative analysis, the experimental results demonstrate that this algorithm could effectively eliminate the effect of uneven illumination on crack detection. After detection, Recall could reach more than 96%, and after extraction, Precision was increased by more than 70%.

show abstract

Autonomous Robotic Inspection in Tunnels

Cited by 22 publications

References 24 publications

Multi-level feature fusion in densely connected deep-learning architecture and depth-first search for crack segmentation on images collected with smartphones

Multi-level feature fusion in densely connected deep-learning architecture and depth-first search for crack segmentation on images collected with smartphones

Deep learning for bridge load capacity estimation in post-disaster and -conflict zones

Dynamic Partition Gaussian Crack Detection Algorithm Based on Projection Curve Distribution

Contact Info

Product

Resources

About