• Proceedings of the Korea Committee for Ocean Resources and Engineering Conference
• /
• 2003.05a
• /
• pp.105-108
• /
• 2003

The local motion at the top of an A-frame fixed on a research vessel for deep sea ROV floating in irregular waves is studied in the time-domain. The motion is analyzed in the time-domain using the convolution integral of the radiation forces. The memory effect functions and infinite frequency added masses are obtained from the solution of the three dimensional improved Green integral equation in the frequency domain by making use of the Fourier transformation.

• Smart Structures and Systems
• /
• v.32 no.2
• /
• pp.111-121
• /
• 2023

This study proposes an automated assembly performance evaluation method for prefabricated steel structures (PSSs) using machine learning methods. Assembly component images were segmented using a modified version of the receptive field pyramid. By factorizing channel modulation and the receptive field exploration layers of the convolution pyramid, highly accurate segmentation results were obtained. After completing segmentation, the positions of the bolt holes were calculated using various image processing techniques, such as fuzzy-based edge detection, Hough's line detection, and image perspective transformation. By calculating the distance ratio between bolt holes, the assembly performance of the PSS was estimated using the k-nearest neighbors (kNN) algorithm. The effectiveness of the proposed framework was validated using a 3D PSS printing model and a field test. The results indicated that this approach could recognize assembly components with an intersection over union (IoU) of 95% and evaluate assembly performance with an error of less than 5%.

• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.940-952
• /
• 2020

The Alzheimer's disease (AD) is a neurodegenerative disease commonly found in the elderly individuals. It is one of the most common forms of dementia; patients with AD suffer from a degradation of cognitive abilities over time. To correctly diagnose AD, compuated-aided system equipped with automatic classification algorithm is of great importance. In this paper, we propose a novel deep learning based classification algorithm that takes advantage of MRI biomarker images including brain areas of hippocampus and cerebrospinal fluid for the purpose of improving the AD classification performance. In particular, we develop a new approach that effectively applies MRI biomarker patch images as input to 3D Deep Convolution Neural Network. To integrate multiple classification results from multiple biomarker patch images, we proposed the effective confidence score fusion that combine classification scores generated from soft-max layer. Experimental results show that AD classification performance can be considerably enhanced by using our proposed approach. Compared to the conventional AD classification approach relying on entire MRI input, our proposed method can improve AD classification performance of up to 10.57% thanks to using biomarker patch images. Moreover, the proposed method can attain better or comparable AD classification performances, compared to state-of-the-art methods.

• Journal of Ocean Engineering and Technology
• /
• v.24 no.5
• /
• pp.16-21
• /
• 2010

During the loading/offloading operation of a liquefied natural gas carrier (LNGC) that is moored in a side-by-side configuration with an offshore plant, sloshing that occurs due to the partially filled LNG tank and the interactive effect between the two floating bodies are important factors that affect safety and operability. Therefore, a time-domain software program, called CHARM3D, was developed to consider the interactions between sloshing and the motion of a floating body, as well as the interactions between multiple bodies using the potential-viscous hybrid method. For the simulation of a floating body in the time domain, hydrodynamic coefficients and wave forces were calculated in the frequency domain using the 3D radiation/diffraction panel program based on potential theory. The calculated values were used for the simulation of a floating body in the time domain by convolution integrals. The liquid sloshing in the inner tanks is solved by the 3D-FDM Navier-Stokes solver that includes the consideration of free-surface non-linearity through the SURF scheme. The computed sloshing forces and moments were fed into the time integration of the ship's motion, and the updated motion was, in turn, used as the excitation force for liquid sloshing, which is repeated for the ensuing time steps. For comparison, a sloshing motion coupled analysis program based on linear potential theory in the frequency domain was developed. The computer programs that were developed were applied to the side-by-side offloading operation between the offshore plant and the LNGC. The frequency-domain results reproduced the coupling effects qualitatively, but, in general, the peaks were over-predicted compared to experimental and time-domain results. The interactive effects between the sloshing liquid and the motion of the vessel can be intensified further in the case of multiple floating bodies.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.35 no.5
• /
• pp.259-266
• /
• 2022

In this study, we developed a deep learning and vision sensor-based assembly performance evaluation method isfor prefabricated steel structures. The assembly parts were segmented using a modified version of the receptive field block convolution module inspired by the eccentric function of the human visual system. The quality of the assembly was evaluated by detecting the bolt holes in the segmented assembly part and calculating the bolt hole positions. To validate the performance of the evaluation, models of standard and defective assembly parts were produced using a 3D printer. The assembly part segmentation network was trained based on the 3D model images captured from a vision sensor. The sbolt hole positions in the segmented assembly image were calculated using image processing techniques, and the assembly performance evaluation using the k-nearest neighbor algorithm was verified. The experimental results show that the assembly parts were segmented with high precision, and the assembly performance based on the positions of the bolt holes in the detected assembly part was evaluated with a classification error of less than 5%.

• Journal of Positioning, Navigation, and Timing
• /
• v.11 no.3
• /
• pp.163-172
• /
• 2022

Typical anti-jamming technologies based on array antennas, Space Time Adaptive Process (STAP) & Space Frequency Adaptive Process (SFAP), are very effective algorithms to perform nulling and beamforming. However, it does not perform equally well for all types of jamming signals. If the anti-jamming algorithm is not optimized for each signal type, anti-jamming performance deteriorates and the operation stability of the system become worse by unnecessary computation. Therefore, jamming classification technique is required to obtain optimal anti-jamming performance. Machine learning, which has recently been in the spotlight, can be considered to classify jamming signal. In general, performing supervised learning for classification requires a huge amount of data and new learning for unfamiliar signal. In the case of jamming signal classification, it is difficult to obtain large amount of data because outdoor jamming signal reception environment is difficult to configure and the signal type of attacker is unknown. Therefore, this paper proposes few-shot jamming signal classification technique using meta-learning and transfer-learning to train the model using a small amount of data. A training dataset is constructed by anti-jamming algorithm input data within the GNSS receiver when jamming signals are applied. For meta-learning, Model-Agnostic Meta-Learning (MAML) algorithm with a general Convolution Neural Networks (CNN) model is used, and the same CNN model is used for transfer-learning. They are trained through episodic training using training datasets on developed our Python-based simulator. The results show both algorithms can be trained with less data and immediately respond to new signal types. Also, the performances of two algorithms are compared to determine which algorithm is more suitable for classifying jamming signals.

• Journal of Korea Multimedia Society
• /
• v.24 no.5
• /
• pp.659-666
• /
• 2021

With the continuous development of deep learning, human behavior recognition algorithms have achieved good results. However, in a multi-person recognition environment, the complex behavior environment poses a great challenge to the efficiency of recognition. To this end, this paper proposes a multi-person pose estimation model. First of all, the human detectors in the top-down framework mostly use the two-stage target detection model, which runs slow down. The single-stage YOLOv3 target detection model is used to effectively improve the running speed and the generalization of the model. Depth separable convolution, which further improves the speed of target detection and improves the model's ability to extract target proposed regions; Secondly, based on the feature pyramid network combined with context semantic information in the pose estimation model, the OHEM algorithm is used to solve difficult key point detection problems, and the accuracy of multi-person pose estimation is improved; Finally, the Euclidean distance is used to calculate the spatial distance between key points, to determine the similarity of postures in the frame, and to eliminate redundant postures.

• Bulletin of the Korean Space Science Society
• /
• 2009.10a
• /
• pp.36.3-37
• /
• 2009

This is an attempt to improve a formula to predict variations of geomagnetic storm indices (Dst) from solar wind parameters. A formula which is most widely accepted was given by Burton et al. (1975) over 30 years ago. Their formula is: dDst*/dt = Q(t) - Dst*(t)/$\tau$, where Q(t) is the Dst injection rate given by the convolution of dawn-to-dusk electric field generated by southward solar wind magnetic field and some response function. However, they did not clearly specify the response function. As a result, misunderstanding seems to be prevailing that the injection rate is proportional to the dawn-to-dusk electric field. In this study we tried to determine the response function by examining 12 intense geomagnetic storms with minimum Dst < -200 nT for which solar wind data are available. The method is as follows. First we assume the form of response function that is specified by several time constants, so that we can calculate the injection rate Q1(t) from the solar wind data. On the other hand, Burton et al. expression provide the observed injection rate Q2(t) = dDst*/dt + Dst*(t)/$\tau$. Thus, it is possible to determine the time constants of response function by a least-squares method to minimize the difference between Q1(t) and Q2(t). We have found this simple method successful enough to reproduce the observed Dst variations from the corresponding solar wind data. The present result provides a scheme to predict the development of Dst 30 minutes to 1 hour in advance by using the real time solar wind data from the ACE spacecraft.

• The Journal of Korean Society for Radiation Therapy
• /
• v.24 no.2
• /
• pp.137-147
• /
• 2012

Purpose: Analyze the Effectiveness of Radiation Treatment Planning by dose calculation and optimization algorithm, apply consideration of actual treatment planning, and then suggest the best way to treatment planning protocol. Materials and Methods: The treatment planning system use Eclipse 10.0. (Varian, USA). PBC (Pencil Beam Convolution) and AAA (Anisotropic Analytical Algorithm) Apply to Dose calculation, DVO (Dose Volume Optimizer 10.0.28) used for optimized algorithm of Intensity Modulated Radiation Therapy (IMRT), PRO II (Progressive Resolution Optimizer V 8.9.17) and PRO III (Progressive Resolution Optimizer V 10.0.28) used for optimized algorithm of VAMT. A phantom for experiment virtually created at treatment planning system, $30{\times}30{\times}30$ cm sized, homogeneous density (HU: 0) and heterogeneous density that inserted air assumed material (HU: -1,000). Apply to clinical treatment planning on the basis of general treatment planning feature analyzed with Phantom planning. Results: In homogeneous density phantom, PBC and AAA show 65.2% PDD (6 MV, 10 cm) both, In heterogeneous density phantom, also show similar PDD value before meet with low density material, but they show different dose curve in air territory, PDD 10 cm showed 75%, 73% each after penetrate phantom. 3D treatment plan in same MU, AAA treatment planning shows low dose at Lung included area. 2D POP treatment plan with 15 MV of cervical vertebral region include trachea and lung area, Conformity Index (ICRU 62) is 0.95 in PBC calculation and 0.93 in AAA. DVO DVH and Dose calculation DVH are showed equal value in IMRT treatment plan. But AAA calculation shows lack of dose compared with DVO result which is satisfactory condition. Optimizing VMAT treatment plans using PRO II obtained results were satisfactory, but lower density area showed lack of dose in dose calculations. PRO III, but optimizing the dose calculation results were similar with optimized the same conditions once more. Conclusion: In this study, do not judge the rightness of the dose calculation algorithm. However, analyzing the characteristics of the dose distribution represented by each algorithm, especially, a method for the optimal treatment plan can be presented when make a treatment plan. by considering optimized algorithm factors of the IMRT or VMAT that needs to optimization make a treatment plan.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.1
• /
• pp.55-61
• /
• 2020

With the development of deep learning technology and advances in computing power, video-based research is now gaining more and more attention. Video data contains a large amount of temporal and spatial information, which is the biggest difference compared with image data. It has a larger amount of data. It has attracted intense attention in computer vision. Among them, motion recognition is one of the research focuses. However, the action recognition of human in the video is extremely complex and challenging subject. Based on many research in human beings, we have found that artificial intelligence-like attention mechanisms are an efficient model for cognition. This efficient model is ideal for processing image information and complex continuous video information. We introduce this attention mechanism into video action recognition, paying attention to human actions in video and effectively improving recognition efficiency. In this paper, we propose a new 3D residual attention network using convolutional neural network based on two attention models to identify human action behavior in the video. An evaluation result of our model showed up to 90.7% accuracy.