• The Journal of Korean Institute of Next Generation Computing
• /
• v.15 no.5
• /
• pp.64-74
• /
• 2019

Although a non-rigid registration has high demands in clinical practice, it has a high computational complexity and it is very difficult for ensuring the accuracy and robustness of registration. This study proposes a method of applying a non-rigid registration to 3D magnetic resonance images of brain in an unsupervised learning environment by using a deep-learning network. A feature vector between two images is produced through the network by receiving both images from two different patients as inputs and it transforms the target image to match the source image by creating a displacement vector field. The network is designed based on a U-Net shape so that feature vectors that consider all global and local differences between two images can be constructed when performing the registration. As a regularization term is added to a loss function, a transformation result similar to that of a real brain movement can be obtained after the application of trilinear interpolation. This method enables a non-rigid registration with a single-pass deformation by only receiving two arbitrary images as inputs through an unsupervised learning. Therefore, it can perform faster than other non-learning-based registration methods that require iterative optimization processes. Our experiment was performed with 3D magnetic resonance images of 50 human brains, and the measurement result of the dice similarity coefficient confirmed an approximately 16% similarity improvement by using our method after the registration. It also showed a similar performance compared with the non-learning-based method, with about 10,000 times speed increase. The proposed method can be used for non-rigid registration of various kinds of medical image data.

• Journal of Korea Water Resources Association
• /
• v.54 no.2
• /
• pp.93-104
• /
• 2021

Due to climate change the sustainable water resources management of agricultural reservoirs, the largest number of reservoirs in Korea, has become important. However, the DIROM, rainfall-runoff model for calculating agricultural reservoir inflow, has used regression equation developed in the 1980s. This study has optimized the parameters of the DIROM using the genetic algorithm (GA) based on historical inflow data for some agricultural reservoirs that recently begun to observe inflow data. The result showed that the error between the historical inflow and simulated inflow using the optimal parameters was decreased by about 80% compared with the annual inflow with the existing parameters. The correlation coefficient and root mean square error with the historical inflow increased to 0.64 and decreased to 28.2 × 103 ㎥, respectively. As a result, if the DIROM uses the optimal parameters based on the historical inflow of agricultural reservoirs, it will be possible to calculate the long-term reservoir inflow with high accuracy. This study will contribute to future research using the historical inflow of agricultural reservoirs and improvement of the rainfall-runoff model parameters. Furthermore, the reliable long-term inflow data will support for sustainable reservoir management and agricultural water supply.

• Journal of the Korean Society of Radiology
• /
• v.15 no.1
• /
• pp.9-14
• /
• 2021

Recently, magnetic resonance imaging (MRI), which can acquire images with good contrast without exposure to radiation, has been widely used for diagnosis. However, noise that reduces the accuracy of diagnosis is essentially generated when acquiring the MR images, and by adjusting the parameters, the noise problem can be solved to obtain an image with excellent characteristics. Among the parameters, the number of excitation (NEX) can acquire images with excellent characteristics without additional degradation of image characteristics. In contrast, appropriate NEX setting is required since the scan time increases and motion artifacts may occur. Therefore, in this study, after fixing all MRI parameters through the MRiLab simulation program, we tried to evaluate the tendency of image characteristics according to changing NEX through quantitative evaluation of brain T2 weighted images acquired by adjusting only NEX. To evaluate the noise level and similarity of the acquired image, signal to noise ratio (SNR), contrast to noise ratio (CNR), root mean square error (RMSE) and peak signal to noise ratio (PSNR) were calculated. As a result, both noise level and similarity evaluation factors showed improved values as NEX increased, while the increasing width gradually decreased. In conclusion, we demonstrated that an appropriate NEX setting is important because an excessively large NEX does not affect image characteristics improvement and cause motion artifacts due to a long scan.

• Journal of Korean Society on Water Environment
• /
• v.37 no.3
• /
• pp.217-228
• /
• 2021

This paper discusses a methodology where an integrated modelling framework is used to quantify the risk derived from anthropic activities on habitats and species. To achieve this purpose, a tool comprising the Delft3D and HABITAT model, was applied in the Yeongsan river. Delft3D effectively simulated the operational condition and flow of weirs in river. In accuracy evaluation of the Delft3D-FLOW, the Bias, Pbias, Mean Absolute Error (MAE), Nash-Sutcliffe Efficiency (NSE), and Index of Agreement (IOA) were used, and the result was evaluated as grade above 'Satisfactory'. The HABITAT calculated Habitat Suitability Value (HSV) for the following eight species: mammal, fish, aquatic plant, and benthic macroinvertebrate. An Area was defined as a suitable habitat if the HSV was larger than 0.5. HABITAT was judged accurately by measuring the Correct Classification rate (CCR) and the area under the ROC curve (AUC). For benthic macroinvertebrate, the CCR and AUC were 77% and 0.834, respectively, at thresholds of 0.017 and 4 inds/m² for HSV and individuals per unit area. This meant that the HABITAT model accurately predicted the appearance of the benthic macroinvertebrates by approximately 77% and that the probability of false alarms was also very low. As a result of evaluating the suitability of habitats, in the Yeongsan river, if the annual "lowest level" (Seungchon weir: 2.5 EL.m/ Juksan weir: -1.35 EL.m) was maintained, the average habitat improvement effect of 6.5%P compared to the 'reference' scenario was predicted. Consequently, it was demonstrated that the integrated modelling framework for habitat suitability assessment is able to support the remedy aquatic ecological management.

• Journal of Food Hygiene and Safety
• /
• v.36 no.1
• /
• pp.34-41
• /
• 2021

The purpose of this research is to improve analysis methods of determining the contents of fatty acids in infant formulas and follow-up formulas. A gas chromatography (GC) method was performed on a GC system coupled to flame ionization detector, with a fused silica capillary column (SP2560, 100 m×0.25 mm, 0.20 ㎛). The method was validated using standard reference material (SRM, NIST 1849a). Performance parameters for method validation such as specificity, linearity, limits of detection (LOD) and quantification (LOQ), accuracy and precision were examined. The linearity of standard solution with correlation coefficient was higher than 0.999 in the range of 0.1-5 mg/mL. The LOD and LOQ were 0.01-0.06 mg/mL and 0.03-0.2 mg/mL, respectively. The recovery using standard reference material was confirmed and the precision was found to be between 0.8% and 2.9% relative standard deviation (RSD). Optimized methods were applied in sample analysis to verify the reliability. All the tested products had acceptable contents of fatty acids compared with component specification for nutrition labeling. The result of this research will provide efficient experimental information and strengthen the management of nutrients in infant formula and follow-up formula.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.6
• /
• pp.573-581
• /
• 2020

Using CCTV images and weather parameters, a method for estimating PM (Particulate Matter) index was proposed, and an experiment was conducted. For CCTV images, we proposed a method of estimating the PM index by applying a deep learning technique based on a CNN (Convolutional Neural Network) with ROI(Region Of Interest) image including a specific spot and an full area image. In addition, after combining the predicted result values by deep learning with the two weather parameters of humidity and wind speed, a post-processing experiment was also conducted to calculate the modified PM index using the learned regression model. As a result of the experiment, the estimated value of the PM index from the CCTV image was R2(R-Squared) 0.58~0.89, and the result of learning the ROI image and the full area image with the measuring device was the best. The result of post-processing using weather parameters did not always show improvement in accuracy in all cases in the experimental area.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.3
• /
• pp.546-559
• /
• 2021

This research aimed to develop the district unit plan simulation using procedural modeling based on shape grammar. For this, Esri's CityEngine 2020.0 was selected as a main development tool, and Inside Commercial Area in Bangi-dong, Songpa-gu, Seoul as the research site where about 25% of the total area was developed over the past five years. Specifically, the research developed the simulation through the following three phases of Data-Information-Knowledge after selecting necessary parameters. In the Data phase, 2 and 3 dimensional data were obtained by utilizing data sharing platforms. In the next Information phase, the acquired data were generated into various procedural models according to the shape grammar, and the 2D and 3D layers were then integrated using relevant applications. In the final Knowledge phase, three-dimensional spatial analysis and storytelling contents were produced based on the integrated layer. As a result, the research suggests the following three implications for the simulation development. First, data accuracy and improvement of sharing platforms are needed in order to effectively carry out the simulation development. Second, the guidelines for district unit plans could be utilized and developed into shape grammar for procedural modeling. Third, procedural modeling is expected to be used as an alternative tool for communication and information delivery.

• Journal of the Korean Institute of Rural Architecture
• /
• v.24 no.1
• /
• pp.9-17
• /
• 2022

Rural spaces are increasingly valuable as areas for introducing renewable energy infrastructure to achieve carbon neutrality. Rural areas are the living grounds of rural residents, and the balance of conservation and development for rural areas is important for the introduction of reasonable facilities. In order to maintain a balance between development and preservation and to introduce reasonable renewable energy facilities, it is necessary to develop a current status survey and an effective survey method to utilize a space capable of introducing renewable energy facilities such as idle land and vacant houses. Therefore, this study was conducted to verify the readability using an unmanned aerial vehicle, and the main results are as follows. The detection of photovoltaic power generation facilities using unmanned aerial vehicles was effective in analyzing the location and area of photovoltaic panels located on the roofs of buildings, and it was possible to calculate the expected power generation by region through the area calculation of photovoltaic panels. The vacant house detection can be used to select an investigation target for an vacant house condition survey as it can identify damage to buildings that are expected to be empty houses, management status, and electricity supply facilities through aerial photos. It is judged that the unmanned aerial vehicle detection capability can be utilized as a method to improve the efficiency of investigation and supplement the data related to solar power generation facilities and vacant houses provided by public institutions. Although this study detected the status of solar power generation facilities and vacant houses through high-resolution aerial image analysis, as a follow-up study, automatic measurement methods using the temperature difference of solar power generation facilities and general characteristics of vacant houses that can be read from the air were investigated. If the deriving research is carried out, it is judged that it will be possible to contribute to the improvement of the accuracy of the detection result using the unmanned aerial vehicle and the expansion of the application range.

• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.101-125
• /
• 2022

Recently, with the development of computing technology and the improvement of the cloud environment, deep learning technology has developed, and attempts to apply deep learning to various fields are increasing. A typical example is anomaly detection, which is a technique for identifying values or patterns that deviate from normal data. Among the representative types of anomaly detection, it is very difficult to detect a contextual anomaly that requires understanding of the overall situation. In general, detection of anomalies in image data is performed using a pre-trained model trained on large data. However, since this pre-trained model was created by focusing on object classification of images, there is a limit to be applied to anomaly detection that needs to understand complex situations created by various objects. Therefore, in this study, we newly propose a two-step pre-trained model for detecting abnormal situation. Our methodology performs additional learning from image captioning to understand not only mere objects but also the complicated situation created by them. Specifically, the proposed methodology transfers knowledge of the pre-trained model that has learned object classification with ImageNet data to the image captioning model, and uses the caption that describes the situation represented by the image. Afterwards, the weight obtained by learning the situational characteristics through images and captions is extracted and fine-tuning is performed to generate an anomaly detection model. To evaluate the performance of the proposed methodology, an anomaly detection experiment was performed on 400 situational images and the experimental results showed that the proposed methodology was superior in terms of anomaly detection accuracy and F1-score compared to the existing traditional pre-trained model.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_2
• /
• pp.1633-1641
• /
• 2022

Urban population concentration and indiscriminate development are causing various environmental problems such as air pollution and heat island phenomena, and causing human resources to deteriorate the damage caused by natural disasters. Urban trees have been proposed as a solution to these urban problems, and actually play an important role, such as providing environmental improvement functions. Accordingly, quantitative measurement and analysis of individual trees in urban trees are required to understand the effect of trees on the urban environment. However, the complexity and diversity of urban trees have a problem of lowering the accuracy of single tree detection. Therefore, we conducted a study to effectively detect trees in Dongjak-gu using high-resolution aerial images that enable effective detection of tree objects and You Only Look Once Version 5 (YOLOv5), which showed excellent performance in object detection. Labeling guidelines for the construction of tree AI learning datasets were generated, and box annotation was performed on Dongjak-gu trees based on this. We tested various scale YOLOv5 models from the constructed dataset and adopted the optimal model to perform more efficient urban tree detection, resulting in significant results of mean Average Precision (mAP) 0.663.