First-Arrival Picking for Microseismic Monitoring Based on Deep Learning
TL;DR: The results show that compared to the U-Net network, the proposed method can obviously improve the first-arrival picking accuracy of the low signal-to-noise ratio microseismic signals, achieving significantly higher accuracy and efficiency than the STA/LTA algorithm, which is famous for its high efficiency in traditional algorithms.
Abstract: In microseismic monitoring, achieving an accurate and efficient first-arrival picking is crucial for improving the accuracy and efficiency of microseismic time-difference source location. In the era of big data, the traditional first-arrival picking method cannot meet the real-time processing requirements of microseismic monitoring process. Using the advanced idea of deep learning-based end-to-end classification and the prominent feature extraction advantages of a fully convolution neural network, this paper proposes a first-arrival picking method of effective signals for microseismic monitoring based on UNet++ network, which can significantly improve the accuracy and efficiency of first-arrival picking. In this paper, we first introduced the methodology of the UNet++-based picking method. And then, the performance of the proposed method is verified by the experiments with finite-difference forward modeling simulated signals and actual microseismic records under different signal-to-noise ratios, and finally, comparative experiments are performed using the U-Net-based first-arrival picking algorithm and the Short-Term Average to Long-Term Average (STA/LTA) algorithm. The results show that compared to the U-Net network, the proposed method can obviously improve the first-arrival picking accuracy of the low signal-to-noise ratio microseismic signals, achieving significantly higher accuracy and efficiency than the STA/LTA algorithm, which is famous for its high efficiency in traditional algorithms.
Citations
More filters
TL;DR: In this paper , the internal relationship between rockburst and microseismic events is analyzed and the occurrence mechanism of slip-type rockburst in underground cavern of Shuangjiangkou Hydropower Station is analyzed through the change of micro seismic monitoring b value to warn the risk level of rockburst.
Abstract: Based on the microseismic monitoring data of underground cavern of Shuangjiangkou Hydropower Station, the internal relationship between rockburst and microseismic events is analyzed and the occurrence mechanism of slip-type rockburst in underground cavern of Shuangjiangkou Hydropower Station is analyzed through the change of microseismic monitoring b value to warn the risk level of rockburst. The results show that microseismic events are closely related to excavation activities, and the change of b value reflects the rockburst risk level. The smaller the b value of microseismic events is, the more large magnitude events is and the greater the possibility of rockburst is. The slip-type rockburst was mainly affected by structural plane; it is due to the accumulation of elastic strain energy at the structural plane and the sudden release under the action of unloading or disturbance, driving the block on the structure surface fast slide, mainly manifested as the sliding of large blocks of surrounding rock, and forming a “V” shape pit blasting or cuneate blasting crater, as well as destruction with crackle. Compared with the strained rockburst, the damage is more serious. The research results can provide reference for rockburst prediction and prevention in similar deep rock engineering.
1 citations
TL;DR: In this article , a segmentation method using convolutional neural network was proposed to improve the accuracy of TOF extraction and further guarantee the accuracy for velocity inversion, and the results demonstrated that the proposed method provided more accurate TOF extractions.
Abstract: Due to the significant acoustic impedance contrast between bone and soft tissue, it is difficult to image bone with conventional ultrasound imaging modalities based on uniform sound velocity. Accurate estimation of the sound velocity distribution is the key basis for bone ultrasonic imaging. The sound velocity model is usually estimated using travel-time inversion based on the extraction of the time-of-flight (TOF). However, TOF extraction is an intractable task for bone imaging, since the first arrival signal is usually of a small amplitude and submerged in noise. To improve the accuracy of TOF extraction and further guarantee the accuracy of velocity inversion, a segmentation method using convolutional neural network was proposed in this work. The TOF extraction was translated into a binary segmentation problem and solved by the data-driven method. After sound velocity inversion, a velocity-corrected version of the Kirchhoff migration method (MKMM) was developed to reconstruct bone image. The performance of the proposed method was verified by numerical bone models and ex-vivo bovine femurs coated with soft-tissue phantoms. The results demonstrated that the proposed method provided more accurate TOF extractions (i.e., the errors < 0.24 μs for numerical models and < 0.4 μs for ex-vivo specimens). The resulting velocity inversion was significantly improved (the mean relative errors < 4.5% for numerical models and < 5.1% for ex-vivo samples). Moreover, the satisfactory reconstructions of bone and the overlying soft tissue can be observed. It is demonstrated that the proposed method is effective for bone ultrasonic imaging.
References
More filters
Proceedings Article•
03 Dec 2012TL;DR: The state-of-the-art performance of CNNs was achieved by Deep Convolutional Neural Networks (DCNNs) as discussed by the authors, which consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax.
Abstract: We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.
73,978 citations
Posted Content•
TL;DR: This paper presents UNet++, a new, more powerful architecture for medical image segmentation where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways, and argues that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar.
Abstract: In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.
2,254 citations
20 Sep 2018
TL;DR: UNet++ as discussed by the authors is a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways.
Abstract: In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.
2,067 citations
TL;DR: Virtual adversarial training (VAT) as discussed by the authors is a regularization method based on virtual adversarial loss, which is a measure of local smoothness of the conditional label distribution given input.
Abstract: We propose a new regularization method based on virtual adversarial loss: a new measure of local smoothness of the conditional label distribution given input. Virtual adversarial loss is defined as the robustness of the conditional label distribution around each input data point against local perturbation. Unlike adversarial training, our method defines the adversarial direction without label information and is hence applicable to semi-supervised learning. Because the directions in which we smooth the model are only “virtually” adversarial, we call our method virtual adversarial training (VAT). The computational cost of VAT is relatively low. For neural networks, the approximated gradient of virtual adversarial loss can be computed with no more than two pairs of forward- and back-propagations. In our experiments, we applied VAT to supervised and semi-supervised learning tasks on multiple benchmark datasets. With a simple enhancement of the algorithm based on the entropy minimization principle, our VAT achieves state-of-the-art performance for semi-supervised learning tasks on SVHN and CIFAR-10.
1,991 citations
TL;DR: UNet++ as mentioned in this paper proposes an efficient ensemble of U-Nets of varying depths, which partially share an encoder and co-learn simultaneously using deep supervision, leading to a highly flexible feature fusion scheme.
Abstract: The state-of-the-art models for medical image segmentation are variants of U-Net and fully convolutional networks (FCN). Despite their success, these models have two limitations: (1) their optimal depth is apriori unknown, requiring extensive architecture search or inefficient ensemble of models of varying depths; and (2) their skip connections impose an unnecessarily restrictive fusion scheme, forcing aggregation only at the same-scale feature maps of the encoder and decoder sub-networks. To overcome these two limitations, we propose UNet++, a new neural architecture for semantic and instance segmentation, by (1) alleviating the unknown network depth with an efficient ensemble of U-Nets of varying depths, which partially share an encoder and co-learn simultaneously using deep supervision; (2) redesigning skip connections to aggregate features of varying semantic scales at the decoder sub-networks, leading to a highly flexible feature fusion scheme; and (3) devising a pruning scheme to accelerate the inference speed of UNet++. We have evaluated UNet++ using six different medical image segmentation datasets, covering multiple imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and electron microscopy (EM), and demonstrating that (1) UNet++ consistently outperforms the baseline models for the task of semantic segmentation across different datasets and backbone architectures; (2) UNet++ enhances segmentation quality of varying-size objects—an improvement over the fixed-depth U-Net; (3) Mask RCNN++ (Mask R-CNN with UNet++ design) outperforms the original Mask R-CNN for the task of instance segmentation; and (4) pruned UNet++ models achieve significant speedup while showing only modest performance degradation. Our implementation and pre-trained models are available at https://github.com/MrGiovanni/UNetPlusPlus .
1,487 citations