Smooth Deep Learning Magnetotelluric Inversion Based on Physics-Informed Swin Transformer and Multiwindow Savitzky–Golay Filter

Liu, Wei; Wang, He; Xi, Zhenzhu; Zhang, Rongqing

doi:10.1109/tgrs.2023.3304313

Cited by 6 publications

(3 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this section, pre-processing using Multi-Window Savitzky-Golay Filter (MWSGF) [29] is discussed. The input gestures data then pre-processed using MWSGF to remove noise and improving the quality of gestures data.…”

Section: B Pre-processing Using Multi-window Savitzky-golay Filtermentioning

confidence: 99%

Product Design Interaction and Experience Based on Virtual Reality Technology

Qianqian Hu

2024

jes

View full text Add to dashboard Cite

Artificial intelligence involves imitating human thought, consciousness, and other aspects. This is similar to having machine with human brain that can think, produce independently. However, unlike human brain, it has a speed, memory advantage. A virtual reality interactive glove is built using a nine-axis inertial sensor, an artificial intelligence deep learning algorithm. In this research work, Product Design Interaction and Experience Based on Virtual Reality Technology (PDIE-VRT-SCGAN) were proposed. Initially, input gestures data are gathered from the Virtual Reality Experiences (VRE) dataset. The input gestures data is then pre-processed using Multi-Window Savitzky-Golay Filter (MWSGF) to reduce noises, increase overall quality of the gestures data. In order to improving overall user engagement in product design interactions on virtual reality (VR) technology, the pre-processed gestures data are then fed into an adversarial network called a Semi-Cycled Generative Adversarial Network (SCGAN). In general, SCGAN does not express some adaption of optimization strategies for determining optimal parameters to promise exact to improving overall user engagement in product design interactions using VR technology. Therefore, FOX-inspired Optimization (FIO) is proposed to enhance weight parameter of SCGAN method, which precisely improving the user experience in product design interaction. The efficacy of PDIE-VRT-SCGAN method is assessed using a number of performance criteria, including tracking accuracy, frame rate, latency, rendering time, error rate, and user error. The proposed PDIE-VRT-SCGAN method attains 22.36%, 25.42% and 18.17% higher tracking accuracy, 21.26%, 15.42% and 19.27% higher latency, 28.36%, 25.32% and 28.27% higher frame rate compared with existing methods, such as design and implementation of virtual reality interactive product software depend on artificial intelligence deep learning algorithm(DVRI-PS-AI-DL), virtual evaluation system for product designing utilizing virtual reality (VES-PD-VR), and analysis of unsatisfying user experiences with unmet psychological needs for virtual reality exergames utilizing deep learning approach (AUUE-UP-VRE-DLA) respectively.

show abstract

Section: B Pre-processing Using Multi-window Savitzky-golay Filtermentioning

confidence: 99%

Product Design Interaction and Experience Based on Virtual Reality Technology

Qianqian Hu

2024

jes

View full text Add to dashboard Cite

show abstract

“…Then, the properly trained PISwin-TUNets were used to invert the three noisy MT datasets. We also reproduced the MWSG (different from ours) smoothing technique newly developed by Liu et al [41], which has demonstrated its effectiveness on noisy MT data, and combined it with our PISwinTUNet, named PISwinTUNet-smooth, for inversion comparison. The inversion results are illustrated in Figure 6.…”

Section: Synthetic Example With Familiar Noisementioning

confidence: 99%

“…where w i represents the ith window with a size of 2i + 1. Following Liu et al [41], the largest window size is set to 65 (L interp /2 + 1), and thus, the corresponding window number m is 31.…”

mentioning

confidence: 99%

Physics-Informed Deep Learning Inversion with Application to Noisy Magnetotelluric Measurements

Liu,

Wang,

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Despite demonstrating exceptional inversion production for synthetic data, the application of deep learning (DL) inversion methods to invert realistic magnetotelluric (MT) measurements, which are inevitably contaminated by noise in acquisition, poses a significant challenge. Hence, to facilitate DL inversion for realistic MT measurements, this work explores developing a noise-robust MT DL inversion method by generating targeted noisy training datasets and constructing a physics-informed neural network. Different from most previous works that only considered the noise of one fixed distribution and level, we propose three noise injection strategies and compare their combinations to mitigate the adverse effect of measurement noise on MT DL inversion results: (1) add synthetic relative noise obeying Gaussian distribution; (2) propose a multiwindow Savitzky–Golay (MWSG) filtering scheme to extract potential and possible noise from the target field data and then introduce them into training data; (3) create an augmented training dataset based on the former two strategies. Moreover, we employ the powerful Swin Transformer as the backbone network to construct a U-shaped DL model (SwinTUNet), based on which a physics-informed SwinTUNet (PISwinTUNet) is implemented to further enhance its generalization ability. In synthetic examples, the proposed noise injection strategies demonstrate impressive inversion effects, regardless of whether they are contaminated by familiar or unfamiliar noise. In a field example, the combination of three strategies drives PISwinTUNet to produce considerably faithful reconstructions for subsurface resistivity structures and outperform the classical deterministic Occam inversions. The experimental results show that the proposed noise-robust DL inversion method based on the noise injection strategies and physics-informed DL architecture holds great promise in processing MT field data.

show abstract

Deep learning transient electromagnetic inversion for seawater intrusion

Wang,

Dai,

Liu

et al. 2024

Journal of Geophysics and Engineering

View full text Add to dashboard Cite

To enhance the capability of transient electromagnetic method(TEM) in detecting seawater intrusion and delineating the boundaries in coastal areas, we developed a deep learning inversion method for TEM data based on the Swin Transformer model in this study. First many standardized resistivity models were designed and generated to describe t the subsurface resistivity structures associated with seawater intrusion in coastal areas .Then, TEM forward modelling was performed to compute the corresponding TEM responses, thereby constructing a seawater intrusion-oriented training dataset. Then, the robust Swin Transformer model was employed as the backbone network to build a deep learning inversion model, named SITEMNet, to derive a direct nonlinear transformation that maps TEM responses to subsurface resistivity models. The proposed SITEMNet inversion technique was validated using simulated data scenarios and actual field TEM measurements, showing great promise in accurately identifying seawater intrusion interface and geological formations.

show abstract

Smooth Deep Learning Magnetotelluric Inversion Based on Physics-Informed Swin Transformer and Multiwindow Savitzky–Golay Filter

Cited by 6 publications

References 42 publications

Product Design Interaction and Experience Based on Virtual Reality Technology

Product Design Interaction and Experience Based on Virtual Reality Technology

Physics-Informed Deep Learning Inversion with Application to Noisy Magnetotelluric Measurements

Deep learning transient electromagnetic inversion for seawater intrusion

Contact Info

Product

Resources

About