Self-organized significance analysis on automatically generated training data for neural networks

Birkenfeld, Sven

doi:10.1109/wac.2014.6935999

Cited by 6 publications

(6 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There is also a vast amount of research that focused on the use of neural networks in processing GPR radargrams. For example, Birkenfeld, S. [75] presented a non-fully connected neural network model that identifies interfering or incomplete hyperbolas. Lei et al [46] proposed the faster region-based convolution neural network (R-CNN) and data augmentation to detect regions of buried objects in GPR images.…”

Section: Terrestrial Mappingmentioning

confidence: 99%

Review of Artificial Intelligence Applications for Virtual Sensing of Underground Utilities

Oguntoye

Laflamme

Sturgill

et al. 2023

Sensors

View full text Add to dashboard Cite

Accurately identifying the location and depth of buried utility assets became a considerable challenge in the construction industry, for which accidental strikes can cause important economic losses and safety concerns. While the collection of as-built utility locations is becoming more accurate, there still exists an important need to be capable of accurately detecting buried utilities in order to eliminate risks associated with digging. Current practices typically involve the use of trained agents to survey and detect underground utilities at locations of interest, which is a costly and time-consuming process. With advances in artificial intelligence (AI), an opportunity arose in conducting virtual sensing of buried utilities by combining robotics (e.g., drones), knowledge, and logic. This paper reviewed methods that are based on AI in mapping underground infrastructure. In particular, the use of AI in aerial and terrestrial mapping of utility assets was reviewed, followed by a summary of AI techniques used in fusing multi-source data in creating underground infrastructure maps. Key observations from the consolidated literature were that (1) when leveraging computer vision methods, automatic mapping techniques vastly focus on manholes localized from aerial imagery; (2) when applied to non-intrusive sensing, AI methods vastly focus on empowering ground-penetrating radar (GPR)-produced data; and (3) data fusion techniques to produce utility maps should be extended to any utility assets/types. Based on these observations, a universal utility mapping model was proposed, one that could enable mapping of underground utilities using limited information available in the form of different sources of data and knowledge.

show abstract

Section: Terrestrial Mappingmentioning

confidence: 99%

Review of Artificial Intelligence Applications for Virtual Sensing of Underground Utilities

Oguntoye

Laflamme

Sturgill

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Automation of the process of detection and extraction of hyperbolas that represent underground objects has been discussed in numerous publications [7][8][9]. Manual analysis of large amounts of data from ground penetrating radar (GPR) is inefficient and time-consuming.…”

Section: Introductionmentioning

confidence: 99%

Automatic classification of underground utilities in Urban Areas: A novel method combining ground penetrating radar and image processing

Pasternak,

Fryskowska-Skibniewska

2024

Archives of Civil Engineering

View full text Add to dashboard Cite

Precise determination of the location of underground utility networks is crucial in the field of civil engineering for: the planning and management of space with densely urbanized areas, infrastructure modernization, during construction and building renovations. In this way, damage to underground utilities can be avoided, damage risks to neighbouring buildings can be minimized, and human and material losses can be prevented. It is important to determine not only the location but also the type of underground utility network. Information about location and network types improves the process of land use design and supports the sustainable development of urban areas, especially in the context of construction works in build-up areas and areas planned for development. The authors were inspired to conduct research on this subject by the development of a methodology for classifying network types based on images obtained in a non-invasive way using a Leica DS2000 ground penetrating radar. The authors have proposed a new classification algorithm based on the geometrical properties of hyperboles that represent underground utility networks. Another aim of the research was to automate the classification process, which may support the user in selecting the type of network in images that are sometimes highly noise-laden. The developed algorithm shortens the time required for image interpretation and the selection of underground objects, which is particularly important for inexperienced operators. The classification results revealed that the average effectiveness of the classification of network types ranged from 42% to 70%, depending on the type of infrastructure.

show abstract

“…Machine learning including object detection is also used in buried pipe detection. For example, there are studies that use neural network models trained on reflected waveform images with appropriate preprocessing (Al-Nuaimy et al, 2000;Birkenfeld, 2010;Gamba & Lossani, 2000;Singh & Nene, 2013). Khudoyarov et al (2020) and Yamaguchi et al (2020Yamaguchi et al ( , 2021 took reflected waveforms as 3D input and used 3D convolutional neural networks to detect buried pipes.…”

Section: Introductionmentioning

confidence: 99%

Iterative application of generative adversarial networks for improved buried pipe detection from images obtained by ground‐penetrating radar

Chun

Kato²

2023

Computer aided Civil Eng

View full text Add to dashboard Cite

Ground‐penetrating radar (GPR) is widely used to determine the location of buried pipes without excavation, and machine learning has been researched to automatically identify the location of buried pipes from the reflected wave images obtained by GPR. In object detection using machine learning, the accuracy of detection is affected by the quantity and quality of training data, so it is important to expand the training data to improve accuracy. This is especially true in the case of buried pipes that are located underground and whose existence cannot be easily confirmed. Therefore, this study developed a method for increasing training data using you only look once v5 (YOLOv5) and StyleGAN2‐ADA to automate the annotation process. Of particular importance is developing a framework for generating images by generative adversarial networks with an emphasis on images that are challenging to detect buried pipes in YOLOv5 and add them to a training dataset to repeat training recursively, which has greatly improved the detection accuracy. Specifically, F‐values of 0.915, 0.916, and 0.924 were achieved by automatically generating training images step by step from only 500, 1000, and 2000 training images, respectively. These values exceed the F‐value of 0.900, which is obtained from training by manually annotating 15,000 images, a much larger number. In addition, we applied the method to a road in Shizuoka Prefecture, Japan, and confirmed that the method can detect the location of buried pipes with high accuracy on a real road. This method can contribute to labor‐saving training data expansion, which is time‐consuming and costly in practice, and as a result, the method contributes to improving detection accuracy.

show abstract

Self-organized significance analysis on automatically generated training data for neural networks

Cited by 6 publications

References 5 publications

Review of Artificial Intelligence Applications for Virtual Sensing of Underground Utilities

Review of Artificial Intelligence Applications for Virtual Sensing of Underground Utilities

Automatic classification of underground utilities in Urban Areas: A novel method combining ground penetrating radar and image processing

Iterative application of generative adversarial networks for improved buried pipe detection from images obtained by ground‐penetrating radar

Contact Info

Product

Resources

About