• Korean Journal of Remote Sensing
• /
• v.37 no.6_1
• /
• pp.1647-1656
• /
• 2021

In order to reduce the spread of infection due to COVID-19, South Korea has established a four-step social distancing standard and implemented it by changing the steps based on the rate of confirmed cases. The implementation of social distancing brought about a change in the amount of activity of citizens by limiting social contact such as movement and gathering of people. One of the data that can intuitively confirm this is Night Time Light (NTL). NTL is a variable that can measure the size of the national economy measured using lights captured by satellites, and can be used to understand people's social activities during the night. The NTL visible data is obtained via the Visible Infrared Imaging Radiometer Suite (VIIRS) Day-Night Band (DNB) onboard the Suomi National Polar-orbiting Partnership (S-NPP) satellite. 1023 of Suomi data from 1 January 2019 until 26 October 2021 were collected to generate time series of NTL radiance change over Seoul to analyze the correlation with social distancing policy. The results show that implementing the level of social distancing generally decreased the NTL radiance both in spatial disparities and temporal patterns. The higher level of policy, limiting human activities combined with the low number of people who have been vaccinated and the closure of various facilities. Because of social distancing, the differences in human activities affected the nighttime light during the COVID-19 pandemic, especially in Seoul, South Korea. Therefore, this study can be used as a reference for the government in evaluating and improving policies related to efforts reducing the transmission of COVID-19.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.417-427
• /
• 2021

Long-range communication in deep waters must overcome the low data rate due to limited bandwidth. This paper presents the performance of Multiple-Input-Multiple-Output (MIMO) system to increase the data rate. In MIMO system, communication performance is degraded by crosstalk between users and an adaptive passive Time Reversal Processing (TRP) is widely used to eliminate this. In October 2018, long-range underwater acoustic communication experiment was conducted in deep water (1,000 m ~) off the east of Pohang, South Korea. During the experiment, a vertical line array was utilized and communication signals modulated by binary phase shift keying and quadrature phase shift keying with a symbol rate of 512 sps were transmitted. To generate MIMO communication signals, received signals from ranges of 26 km and 30 km is synthesized. Compared to the conventional passive TRP, the adaptive passive TRP eliminates the crosstalk between users and achieves error-free performance with an increase of output signal-to-noise ratio. Therefore, two users separated by 4 km in range achieves an aggregate data rate of 1,024 symbols/s.

• Journal of Korea Water Resources Association
• /
• v.54 no.10
• /
• pp.779-793
• /
• 2021

Water stress and environmental drivers are important factors to explain the variance of gross primary production (GPP). Environmental drivers are used to generate GPP in Moderate Resolution Imaging Spectroradiometer (MODIS) algorithm and process-based model. However, MODIS algorithm only consider the vapor pressure deficit (VPD) data while the process-based biogeochemical model also uses limited data to express water stress. We compared the relationship between environmental drivers and GPP from eddy covariance method, MODIS algorithm, and Community Land Model 4 (CLM 4) simulation in normal years and drought years. To consider water stress specifically, we used VPD and evaporative fraction (EF). We evaluated the effects from environmental drivers and EF towards GPP products using the structural equation modeling (SEM) in South Korea. We found that GPP products from MODIS algorithm and model simulation results were not restricted from VPD data if VPD was underestimated. We also found that in the cropland area, irrigation effects can relieve VPD effects to GPP. However, GPP products derived from MODIS and CLM 4 had limitation to explain the irrigation effects to GPP. Overall, these results will enhance the understanding of GPP products derived from MODIS and CLM 4.

• Korean Journal of Remote Sensing
• /
• v.37 no.6_3
• /
• pp.1931-1942
• /
• 2021

As the utility of an optical satellite image with a high spatial resolution (i.e., fine-scale) has been emphasized, recently, various studies of the land surface monitoring using those have been widely carried out. However, the usefulness of fine-scale satellite images is limited because those are acquired at a low temporal resolution. To compensate for this limitation, the spatiotemporal data fusion can be applied to generate a synthetic image with a high spatio-temporal resolution by fusing multiple satellite images with different spatial and temporal resolutions. Since the spatio-temporal data fusion models have been developed for mid or low spatial resolution satellite images in the previous studies, it is necessary to evaluate the applicability of the developed models to the satellite images with a high spatial resolution. For this, this study evaluated the applicability of the developed spatio-temporal fusion models for KOMPSAT-3A and Sentinel-2 images. Here, an Enhanced Spatial and Temporal Adaptive Fusion Model (ESTARFM) and Spatial Time-series Geostatistical Deconvolution/Fusion Model (STGDFM), which use the different information for prediction, were applied. As a result of this study, it was found that the prediction performance of STGDFM, which combines temporally continuous reflectance values, was better than that of ESTARFM. Particularly, the prediction performance of STGDFM was significantly improved when it is difficult to simultaneously acquire KOMPSAT and Sentinel-2 images at a same date due to the low temporal resolution of KOMPSAT images. From the results of this study, it was confirmed that STGDFM, which has relatively better prediction performance by combining continuous temporal information, can compensate for the limitation to the low revisit time of fine-scale satellite images.

• Journal of Food Hygiene and Safety
• /
• v.34 no.2
• /
• pp.199-204
• /
• 2019

This study aimed to generate a probability distribution model based on temperature data of frozen food storage facility as input variables for microbial risk assessment (MRA). We visited 8 food-handling businesses to collect temperature data from their cold storage warehouses. The overall mean temperature inside the storage facilities was $-20.48{\pm}3.08^{\circ}C$, with 20.4% of the facilities having above $-18^{\circ}C$, with minimum and maximum temperature values of -10.3 and $-25.80^{\circ}C$ respectively. Temperature distributions by space locations of natural and forced convection were $-22.57{\pm}0.84$ and $-17.81{\pm}1.47^{\circ}C$, $-22.49{\pm}1.05$ and $-17.94{\pm}1.44^{\circ}C$, and $-22.68{\pm}1.03$ and $-18.08{\pm}1.42^{\circ}C$ in the upper (2.4~4 m), middle (1.5~2.4 m), and lower (0.7~1.5 m) shelves, respectively. Probability distributions from the temperature data were obtained using the program @RISK. Statistical ranking was determined using goodness of fit to determine the probability distribution model. Our results show that a log-normal distribution [5.9731, 3.3483, shift (-26.4281)] is most appropriate for relative MRA conduction.

• Journal of Korea Water Resources Association
• /
• v.54 no.5
• /
• pp.301-309
• /
• 2021

In this study, we tried to improve the performance of the existing U-net-based deep learning rainfall prediction model, which can weaken the meaning of time series order. For this, ConvLSTM2D U-Net structure model considering temporal consistency of data was applied, and we evaluated accuracy of the ConvLSTM2D U-Net model using a RainNet model and an extrapolation-based advection model. In addition, we tried to improve the uncertainty in the model training process by performing learning not only with a single model but also with 10 ensemble models. The trained neural network rainfall prediction model was optimized to generate 10-minute advance prediction data using four consecutive data of the past 30 minutes from the present. The results of deep learning rainfall prediction models are difficult to identify schematically distinct differences, but with ConvLSTM2D U-Net, the magnitude of the prediction error is the smallest and the location of rainfall is relatively accurate. In particular, the ensemble ConvLSTM2D U-Net showed high CSI, low MAE, and a narrow error range, and predicted rainfall more accurately and stable prediction performance than other models. However, the prediction performance for a specific point was very low compared to the prediction performance for the entire area, and the deep learning rainfall prediction model also had limitations. Through this study, it was confirmed that the ConvLSTM2D U-Net neural network structure to account for the change of time could increase the prediction accuracy, but there is still a limitation of the convolution deep neural network model due to spatial smoothing in the strong rainfall region or detailed rainfall prediction.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.15 no.1
• /
• pp.56-61
• /
• 2022

In this paper, we classify and present AI algorithms and natural language processing methods used in chatbots. A framework that can be used to implement a chatbot is also described. A chatbot is a system with a structure that interprets the input string by constructing the user interface in a conversational manner and selects an appropriate answer to the input string from the learned data and outputs it. However, training is required to generate an appropriate set of answers to a question and hardware with considerable computational power is required. Therefore, there is a limit to the practice of not only developing companies but also students learning AI development. Currently, chatbots are replacing the existing traditional tasks, and a practice course to understand and implement the system is required. RNN and Char-CNN are used to increase the accuracy of answering questions by learning unstructured data by applying technologies such as deep learning beyond the level of responding only to standardized data. In order to implement a chatbot, it is necessary to understand such a theory. In addition, the students presented examples of implementation of the entire system by utilizing the methods that can be used for coding education and the platform where existing developers and students can implement chatbots.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.12
• /
• pp.579-586
• /
• 2021

Time series forecasting refers to predicting future time information based on past time information. Accurately predicting future information is crucial because it is used for establishing strategies or making policy decisions in various fields. Recently, a transformer model has been mainly studied for a time series prediction model. However, the existing transformer model has a limitation in that it has an auto-regressive structure in which the output result is input again when the prediction sequence is output. This limitation causes a problem in that accuracy is lowered when predicting a distant time point. This paper proposes a sequential decoding model focusing on the style transformation technique to handle these problems and make more precise time series forecasting. The proposed model has a structure in which the contents of past data are extracted from the transformer-encoder and reflected in the style-based decoder to generate the predictive sequence. Unlike the decoder structure of the conventional auto-regressive transformer, this structure has the advantage of being able to more accurately predict information from a distant view because the prediction sequence is output all at once. As a result of conducting a prediction experiment with various time series datasets with different data characteristics, it was shown that the model presented in this paper has better prediction accuracy than other existing time series prediction models.

• Korean Journal of Remote Sensing
• /
• v.38 no.2
• /
• pp.199-213
• /
• 2022

The unsupervised domain adaptation can solve the impractical issue of repeatedly collecting high-quality training data every year for annual crop classification. This study evaluates the applicability of deep learning-based unsupervised domain adaptation models for crop classification. Three unsupervised domain adaptation models including a deep adaptation network (DAN), a deep reconstruction-classification network, and a domain adversarial neural network (DANN) are quantitatively compared via a crop classification experiment using unmanned aerial vehicle images in Hapcheon-gun and Changnyeong-gun, the major garlic and onion cultivation areas in Korea. As source baseline and target baseline models, convolutional neural networks (CNNs) are additionally applied to evaluate the classification performance of the unsupervised domain adaptation models. The three unsupervised domain adaptation models outperformed the source baseline CNN, but the different classification performances were observed depending on the degree of inconsistency between data distributions in source and target images. The classification accuracy of DAN was higher than that of the other two models when the inconsistency between source and target images was low, whereas DANN has the best classification performance when the inconsistency between source and target images was high. Therefore, the extent to which data distributions of the source and target images match should be considered to select the best unsupervised domain adaptation model to generate reliable classification results.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.50 no.8
• /
• pp.573-581
• /
• 2022

This paper deals with the development of CanSat system, which ejects two maple seed-type autorotating science payloads and collects weather information. The CanSat consists of two autorotating science payloads and a container. The container is equipped with devices for launching science payloads and communication with the ground station, and launches science payloads one by one at different designated altitudes. The science payload consists of a space for loading and a large wing, and rotates to generate lift for slowing down the fall speed. Specifically, after being ejected, it descends at a speed of 20 m/s or less, measures the rotation rate, atmospheric pressure, and temperature, and transmits the measured value to the container at a rate of once per second. The communication system is a master-slave structure, and the science payload transmits all data to the master container, which aggregates both the received data and its own data, and transmits it to the ground station. All telemetry can be checked in real time using the ground station software developed in-house. A simulation was performed in the simulation environment, and the performance of the CanSat system that satisfies the mission requirements was confirmed.