• Applied Microscopy
• /
• v.50
• /
• pp.23.1-23.9
• /
• 2020

We propose an effective deep learning model to denoise scanning transmission electron microscopy (STEM) image series, named Noise2Atom, to map images from a source domain 𝓢 to a target domain 𝓒, where 𝓢 is for our noisy experimental dataset, and 𝓒 is for the desired clear atomic images. Noise2Atom uses two external networks to apply additional constraints from the domain knowledge. This model requires no signal prior, no noise model estimation, and no paired training images. The only assumption is that the inputs are acquired with identical experimental configurations. To evaluate the restoration performance of our model, as it is impossible to obtain ground truth for our experimental dataset, we propose consecutive structural similarity (CSS) for image quality assessment, based on the fact that the structures remain much the same as the previous frame(s) within small scan intervals. We demonstrate the superiority of our model by providing evaluation in terms of CSS and visual quality on different experimental datasets.

• Journal of KIISE:Software and Applications
• /
• v.26 no.8
• /
• pp.1000-1009
• /
• 1999

본 논문에서는 비디오 데이타를 분석하여 다양한 카메라의 동작을 정량적으로 추출하는 방법을 제안한다. 본 논문에서 제안하는 카메라의 동작 추출 방법은 어파인 모델을 이용한 방법으로 인접 영상으로부터 추출한 동작 벡터를 어파인 모델에 적용하고 회귀분석법을 통해 어파인 모델을 구성하는 파라미터를 구한다. 그런 다음, 파라미터들의 크기를 분석하고 상호 관계를 해석하여 카메라의 동작을 추출한다. 본 논문에서는 잡음이 포함된 동작 벡터를 필터링하여 사용하므로 잡음에 강건한 결과를 얻을 수 있다. 그리고 어파인 모델을 구성하는 파라미터만을 분석함으로써 카메라의 다양한 동작을 간단하면서도 비교적 정확하게 추출한다. 실험 결과는 카메라의 동작을 정확하게 추출하고 있음을 보여준다.Abstract This paper presents an elegant method, an affine-model based approach, that can qualitatively estimate the information of camera motion. We define various types of camera motion by means of parameters of an affine-model. To get those parameters from images, we fit an affine-model to the field of instantaneous velocities, rather than raw images. We correlate consecutive images to get instantaneous velocities. The size filtering of the velocities are applied to remove noisy components, and the regression approach is employed for the fitting procedure. The fitted values of the parameters are examined to get the estimates of camera motion. The experimental results show that the suggested approach can yield the qualitative information of camera motion successfully.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.1
• /
• pp.64-68
• /
• 2020

In a conventional speech recognition system using Gaussian Mixture Model-Hidden Markov Model (GMM-HMM), the cepstral feature normalization method based on pole filtering was effective in improving the performance of recognition of short utterances in noisy environments. In this paper, the usefulness of this method for the state-of-the-art speech recognition system using Deep Neural Network (DNN) is examined. Experimental results on AURORA 2 DB show that the cepstral mean and variance normalization based on pole filtering improves the recognition performance of very short utterances compared to that without pole filtering, especially when there is a large mismatch between the training and test conditions.

• Journal of Power Electronics
• /
• v.15 no.5
• /
• pp.1274-1285
• /
• 2015

This paper deals with model reference adaptive system (MRAS) speed estimators based on a secondary flux for linear induction motors (LIMs). The operation of these estimators significantly depends on an adaptation mechanism. Fixed-gain PI controller is the most common adaptation mechanism that may fail to estimate the speed correctly in different conditions, such as variation in machine parameters and noisy environment. Two adaptation mechanisms are proposed to improve LIM drive system performance, particularly at very low speed. The first adaptation mechanism is based on fuzzy theory, and the second is obtained from an LIM mechanical model. Compared with a conventional PI controller, the proposed adaptation mechanisms have low sensitivity to both variations of machine parameters and noise. The optimum parameters of adaptation mechanisms are tuned using an offline method through chaotic optimization algorithm (COA) because no design criterion is given to provide these values. The efficiency of MRAS speed estimator is validated by both numerical simulation and real-time hardware-in-the-loop (HIL) implementations. Results indicate that the proposed adaptation mechanisms improve performance of MRAS speed estimator.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.7C
• /
• pp.937-945
• /
• 2004

A binary decision model which is used to denoising has demerits to measure the precise ratio of signal to noise because of only a binary classification. To supplement these demerits, complex statistical model and undecimated wavelet transform are generally exploited. In this paper, we propose a noise reduction method using a multi-level decision model for measuring the ratio of noise in noisy image. The propose method achieves good denoising performance with orthogonal wavelet transform because the ratio of signal to noise can be calculated to multi-valued form. In simulation results, the proposed denoising method outperforms 0.1dB in the PSNR sense than the state of art denoising algorithms using orthogonal wavelet transform.

• Nuclear Engineering and Technology
• /
• v.46 no.6
• /
• pp.773-782
• /
• 2014

The general path model (GPM) is one approach for performing degradation-based, or Type III, prognostics. The GPM fits a parametric function to the collected observations of a prognostic parameter and extrapolates the fit to a failure threshold. This approach has been successfully applied to a variety of systems when a sufficient number of prognostic parameter observations are available. However, the parametric fit can suffer significantly when few data are available or the data are very noisy. In these instances, it is beneficial to include additional information to influence the fit to conform to a prior belief about the evolution of system degradation. Bayesian statistical approaches have been proposed to include prior information in the form of distributions of expected model parameters. This requires a number of run-to-failure cases with tracked prognostic parameters; these data may not be readily available for many systems. Reliability information and stressor-based (Type I and Type II, respectively) prognostic estimates can provide the necessary prior belief for the GPM. This article presents the Bayesian updating framework to include prior information in the GPM and compares the efficacy of including different information sources on two data sets.

• Computational Structural Engineering : An International Journal
• /
• v.3 no.1
• /
• pp.1-7
• /
• 2003

For system identification of large structures, it is not practical to identify the entire structure due to the prohibitive computational time and difficulty in numerical convergence. This paper explores the possibility of performing system identification at substructure level, taking advantage of reduction in both the number of unknowns and the number of degrees of freedom involved. Another advantage is that different portions (substructures) of a structural system can be identified independently and even concurrently with parallel computing. Two substructural identification methods are formulated on the basis whether substructural approach is used to obtain first-order or second-order model. For substructural first-order model, identification at the substructure level will be performed by means of the Observer/Kalman filter Identification (OKID) and the Eigensystem Realization Algorithm (ERA) whereas identification at the global level will be performed to obtain second-order model in order to evaluate the system's stiffness and mass parameters. In the case of substructural second-order model, identification will be performed at the substructure level throughout the identification process. The efficiency of the proposed technique is shown by numerical examples for multi-storey shear buildings subjected to random forces, taking into consideration the effects of noisy measurement data. The results indicate that both the proposed methods are effective and efficient for damage identification of large structures.

• Journal of Information Processing Systems
• /
• v.13 no.5
• /
• pp.1168-1182
• /
• 2017

In the past decades, various image regularization methods have been introduced. Among them, total variation model has drawn much attention for the reason of its low computational complexity and well-understood mathematical behavior. However, regularization parameter estimation of total variation model is still an open problem. To deal with this problem, a novel adaptive regularization parameter selection scheme is proposed in this paper, by means of using the local spectral response, which has the capability of locally selecting the regularization parameters in a content-aware way and therefore adaptively adjusting the weights between the two terms of the total variation model. Experiment results on simulated and real noisy image show the good performance of our proposed method, in visual improvement and peak signal to noise ratio value.

• Structural Engineering and Mechanics
• /
• v.82 no.2
• /
• pp.245-258
• /
• 2022

Many Structural Health Monitoring (SHM) methods have been proposed for structural damage diagnosis and prognosis. However, SHM for pinched hysteretic structures can be problematic due to the high level of nonlinearity. The model-free hysteresis loop analysis (HLA) has displayed notable robustness and accuracy in identifying damage for full-scaled and scaled test buildings. In this paper, the performance of HLA is compared with seven other SHM methods in identifying lateral elastic stiffness for a six-story numerical building with highly nonlinear pinching behavior. Two successive earthquakes are employed to compare the accuracy and consistency of methods within and between events. Robustness is assessed across sampling rates 50-1000 Hz in noise-free condition and then assessed with 10% root mean square (RMS) noise added to responses at 250 Hz sampling rate. Results confirm HLA is the most robust method to sampling rate and noise. HLA preserves high accuracy even when the sampling rate drops to 50 Hz, where the performance of other methods deteriorates considerably. In noisy conditions, the maximum absolute estimation error is less than 4% for HLA. The overall results show HLA has high robustness and accuracy for an extremely nonlinear, but realistic case compared to a range of leading and recent model-based and model-free methods.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.104-107
• /
• 2022

autoencoder deep learning model has excellent ability to restore abnormal data to normal data, so it is not appropriate for anomaly detection. In addition, the Inpainting method, which is a method of restoring hidden data after masking (masking) a part of the data, has a problem in that the restoring ability is poor for noisy images. In this paper, we use a method of modifying and improving the MLP-Mixer model to mask the image at a certain ratio and to reconstruct the image by delivering compressed information of the masked image to the model. After constructing a model learned with normal data from the MVTec AD dataset, a reconstruction error was obtained by inputting normal and abnormal images, respectively, and anomaly detection was performed through this. As a result of the performance evaluation, it was found that the proposed method has superior anomaly detection performance compared to the existing method.