• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.175-197
• /
• 2021

Recently, with the development of deep learning technology, research on unstructured data analysis is being actively conducted, and it is showing remarkable results in various fields such as classification, summary, and generation. Among various text analysis fields, text classification is the most widely used technology in academia and industry. Text classification includes binary class classification with one label among two classes, multi-class classification with one label among several classes, and multi-label classification with multiple labels among several classes. In particular, multi-label classification requires a different training method from binary class classification and multi-class classification because of the characteristic of having multiple labels. In addition, since the number of labels to be predicted increases as the number of labels and classes increases, there is a limitation in that performance improvement is difficult due to an increase in prediction difficulty. To overcome these limitations, (i) compressing the initially given high-dimensional label space into a low-dimensional latent label space, (ii) after performing training to predict the compressed label, (iii) restoring the predicted label to the high-dimensional original label space, research on label embedding is being actively conducted. Typical label embedding techniques include Principal Label Space Transformation (PLST), Multi-Label Classification via Boolean Matrix Decomposition (MLC-BMaD), and Bayesian Multi-Label Compressed Sensing (BML-CS). However, since these techniques consider only the linear relationship between labels or compress the labels by random transformation, it is difficult to understand the non-linear relationship between labels, so there is a limitation in that it is not possible to create a latent label space sufficiently containing the information of the original label. Recently, there have been increasing attempts to improve performance by applying deep learning technology to label embedding. Label embedding using an autoencoder, a deep learning model that is effective for data compression and restoration, is representative. However, the traditional autoencoder-based label embedding has a limitation in that a large amount of information loss occurs when compressing a high-dimensional label space having a myriad of classes into a low-dimensional latent label space. This can be found in the gradient loss problem that occurs in the backpropagation process of learning. To solve this problem, skip connection was devised, and by adding the input of the layer to the output to prevent gradient loss during backpropagation, efficient learning is possible even when the layer is deep. Skip connection is mainly used for image feature extraction in convolutional neural networks, but studies using skip connection in autoencoder or label embedding process are still lacking. Therefore, in this study, we propose an autoencoder-based label embedding methodology in which skip connections are added to each of the encoder and decoder to form a low-dimensional latent label space that reflects the information of the high-dimensional label space well. In addition, the proposed methodology was applied to actual paper keywords to derive the high-dimensional keyword label space and the low-dimensional latent label space. Using this, we conducted an experiment to predict the compressed keyword vector existing in the latent label space from the paper abstract and to evaluate the multi-label classification by restoring the predicted keyword vector back to the original label space. As a result, the accuracy, precision, recall, and F1 score used as performance indicators showed far superior performance in multi-label classification based on the proposed methodology compared to traditional multi-label classification methods. This can be seen that the low-dimensional latent label space derived through the proposed methodology well reflected the information of the high-dimensional label space, which ultimately led to the improvement of the performance of the multi-label classification itself. In addition, the utility of the proposed methodology was identified by comparing the performance of the proposed methodology according to the domain characteristics and the number of dimensions of the latent label space.

• Journal of Korea Water Resources Association
• /
• v.56 no.2
• /
• pp.75-89
• /
• 2023

As the Mekong River basin is a nationally shared river, it is difficult to collect precipitation data, and the quantitative and qualitative quality of the data sets differs from country to country, which may increase the uncertainty of hydrological analysis results. Recently, with the development of remote sensing technology, it has become easier to obtain grid-based precipitation products(GPPs), and various hydrological analysis studies have been conducted in unmeasured or large watersheds using GPPs. In this study, rainfall-runoff simulation in the Mekong River basin was conducted using the SWAT model, which is a quasi-distribution model with three satellite GPPs (TRMM, GSMaP, PERSIANN-CDR) and two GPPs (APHRODITE, GPCC). Four water level stations, Luang Prabang, Pakse, Stung Treng, and Kratie, which are major outlets of the main Mekong River, were selected, and the parameters of the SWAT model were calibrated using APHRODITE as an observation value for the period from 2001 to 2011 and runoff simulations were verified for the period form 2012 to 2013. In addition, using the ConvAE, a convolutional neural network model, spatio-temporal correction of original satellite precipitation products was performed, and rainfall-runoff performances were compared before and after correction of satellite precipitation products. The original satellite precipitation products and GPCC showed a quantitatively under- or over-estimated or spatially very different pattern compared to APHPRODITE, whereas, in the case of satellite precipitation prodcuts corrected using ConvAE, spatial correlation was dramatically improved. In the case of runoff simulation, the runoff simulation results using the satellite precipitation products corrected by ConvAE for all the outlets have significantly improved accuracy than the runoff results using original satellite precipitation products. Therefore, the bias correction technique using the ConvAE technique presented in this study can be applied in various hydrological analysis for large watersheds where rain guage network is not dense.

• Journal of Cadastre & Land InformatiX
• /
• v.52 no.2
• /
• pp.171-187
• /
• 2022

Recently, the demand and development for non-face-to-face services are rapidly progressing due to the pandemic caused by the COVID-19, and attention is focused on the metaverse at the center. Entering the era of the 4th industrial revolution, Metaverse, which means a world beyond virtual and reality, combines various sensing technologies and 3D reconstruction technologies to provide various information and services to users easily and quickly. In particular, due to the miniaturization and economic increase of convergence sensors such as unmanned aerial vehicle(UAV) capable of high-resolution imaging and high-precision LiDAR(Light Detection and Ranging) sensors, research on digital-Twin is actively underway to create and simulate real-life twins. In addition, Game engines in the field of computer graphics are developing into metaverse engines by expanding strong 3D graphics reconstuction and simulation based on dynamic operations. This study constructed a mirror-world type metaverse that reflects real-world coordinate-based reality using Unreal Engine 5, a recently announced metaverse engine, with accurate 3D spatial information data of convergence sensors based on unmanned aerial system(UAS) and LiDAR. and then, spatial information contents and simulations for users were produced based on various public data to verify the accuracy of reconstruction, and through this, it was possible to confirm the construction of a more realistic and highly utilizable metaverse. In addition, when constructing a metaverse that users can intuitively and easily access through the unreal engine, various contents utilization and effectiveness could be confirmed through coordinate-based 3D spatial information with high reproducibility.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.23 no.4
• /
• pp.329-339
• /
• 2021

Soybeans (Glycine max), one of major upland crops, require precise management of environmental conditions, such as temperature, water, and soil, during cultivation since they are sensitive to environmental changes. Application of spectral technologies that measure the physiological state of crops remotely has great potential for improving quality and productivity of the soybean by estimating yields, physiological stresses, and diseases. In this study, we developed and validated a soybean growth prediction model using multispectral imagery. We conducted a linear regression analysis between vegetation indices and soybean growth data (fresh weight and LAI) obtained at Miryang fields. The linear regression model was validated at Goesan fields. It was found that the model based on green ratio vegetation index (GRVI) had the greatest performance in prediction of fresh weight at the calibration stage (R²=0.74, RMSE=246 g/m², RE=34.2%). In the validation stage, RMSE and RE of the model were 392 g/m² and 32%, respectively. The errors of the model differed by cropping system, For example, RMSE and RE of model in single crop fields were 315 g/m² and 26%, respectively. On the other hand, the model had greater values of RMSE (381 g/m²) and RE (31%) in double crop fields. As a result of developing models for predicting a fresh weight into two years (2018+2020) with similar accumulated temperature (AT) in three years and a single year (2019) that was different from that AT, the prediction performance of a single year model was better than a two years model. Consequently, compared with those models divided by AT and a three years model, RMSE of a single crop fields were improved by about 29.1%. However, those of double crop fields decreased by about 19.6%. When environmental factors are used along with, spectral data, the reliability of soybean growth prediction can be achieved various environmental conditions.

• Journal of the Korean earth science society
• /
• v.27 no.4
• /
• pp.450-463
• /
• 2006

Seasonal threshold values for fog detection over the ten airport areas within the Korean Peninsula have been derived from the data of polar-orbit Aqua/Terra MODIS and geostationary GOES-9 during a two years. The values are obtained from reflectance at $0.65{\mu}m\;(R_{0.65})$ and the difference in brightness temperature between $3.7{\mu}m\;and\;11{\mu}m\;(T_{3.7-11})$. In order to examine the discrepancy between the threshold values of two kinds of satellites, the following four parameters have been analyzed under the condition of daytime/nighttime and fog/clear-sky, utilizing their simultaneous observations over the Seoul metropolitan area: brightness temperature at $3.7{\mu}m$, the temperature at $11{\mu}m,\;the\;T_{3.7-11}$ for day and night, and the $R_{0.65}$ for daytime. The parameters show significant correlations (r<0.5) in spatial distribution between the two kinds of satellites. The discrepancy between their infrared thresholds is mainly due to the disagreement in their spatial resolutions and spectral bands, particularly at $3.7{\mu}m$. Fog detection from GOES-9 over the nine airport areas except the Cheongju airport has revealed accuracy of 60% in the daytime and 70% in the nighttime, based on statistical verification. The accuracy decreases in foggy cases with twilight, precipitation, short persistence, or the higher cloud above fog. The sensitivity of radiance and reflectance with wavelength has been analyzed in numerical experiments with respect to various meteorological conditions to investigate optical characteristics of the three channels.

• Journal of Korean Society of Forest Science
• /
• v.106 no.3
• /
• pp.310-319
• /
• 2017

This study categorized the damaged trees by Supervised Classification using time-series-aerial photographs of Bukhan, Cheonggae and Suri mountains because oak wilt disease seemed to be concentrated in the metropolitan regions. In order to analyze the spatial characteristics of the damaged areas, the geographical characteristics such as elevation and slope were statistically analyzed to confirm their strong correlation. Based on the results from the statistical analysis of Moran's I, we have retrieved the following: (i) the value of Moran's I in Bukhan mountain is estimated to be 0.25, 0.32, and 0.24 in 2009, 2010 and 2012, respectively. (ii) the value of Moran's I in Cheonggye mountain estimated to be 0.26, 0.32 and 0.22 in 2010, 2012 and 2014, respectively and (iii) the value of Moran's I in Suri mountain estimated to be 0.42 and 0.42 in 2012 and 2014. respectively. These numbers suggest that the damaged trees are distributed in clusters. In addition, we conducted hotspot analysis to identify how the damaged tree clusters shift over time and we were able to verify that hotspots move in time series. According to our research outcome from the analysis of the entire hotspot areas (z-score>1.65), there were 80 percent probability of oak wilt disease occurring in the broadleaf or mixed-stand forests with elevation of 200~400 m and slope of 20~40 degrees. This result indicates that oak wilt disease hotspots can occur or shift into areas with the above geographical features or forest conditions. Therefore, this research outcome can be used as a basic resource when predicting the oak wilt disease spread-patterns, and it can also prevent disease and insect pest related harms to assist the policy makers to better implement the necessary solutions.

• Journal of the Korean Society for Marine Environment & Energy
• /
• v.18 no.4
• /
• pp.304-309
• /
• 2015

To evaluate the degree of contamination caused by oil spill accident in the sea, the in-situ sensors which are based on the scientific method are needed in the real site. The sensors which are based on the fluorescence detection theory can provide the useful data, such as the concentration of oil. However these kinds of sensors commonly are composed of the ultraviolet (UV) light source such as UV mercury lamp, the multiple excitation/emission filters and the optical sensor which is mainly photomultiplier tube (PMT) type. Therefore, the size of the total sensing platform is large not suitable to be handled in the oil spill field and also the total price of it is extremely expensive. To overcome these drawbacks, we designed the fluorimeter for the oil spill detection which has compact size and cost effectiveness. Before the detail design process, we conducted the experiments to measure the excitation and emission spectrum of oils using five different kinds of crude oils and three different kinds of processed oils. And the fluorescence spectrometer were used to analyze the excitation and emission spectrum of oil samples. We have compared the spectrum results and drawn the each common spectrum regions of excitation and emission. In the experiments, we can see that the average gap between maximum excitation and emission peak wavelengths is near 50 nm for the every case. In the experiment which were fixed by the excitation wavelength of 365 nm and 405 nm, we can find out that the intensity of emission was weaker than that of 280 nm and 325 nm. So, if the light sources having the wavelength of 365 nm or 405 nm are used in the design process of fluorimeter, the optical sensor needs to have the sensitivity which can cover the weak light intensity. Through the results which were derived by the experiment, we can define the important factors which can be useful to select the effective wavelengths of light source, photo detector and filters.

• Journal of the Korean Institute of Landscape Architecture
• /
• v.40 no.5
• /
• pp.8-18
• /
• 2012

Urban nature plays an important role not only in the improvement of the physical environment but also from the perspective of psychological and social function. In particular, schoolyards as well as the green spaces near school surroundings function as a primary space for urban children to experience nature in Korea, as they spend most of their time at school. In this study, the status of urban green spaces near school surroundings was examined. For the analysis, 185 elementary schools in Seoul were selected and the green spaces within a radius of 300m(defined as 'school zone' in this study) were analyzed using the Rapid Eye multispectral satellite image data. The mean green space ratio of school zone accounts to about 21% with a high variation from 74% to 0.7% and more than half of the school zone have a green space ratio of less than 20%. Schools with a high green space ratio in their school zone are mostly located near urban forests, so forest areas particularly contribute to increase the green space ratio. Furthermore, forest vegetation shows relatively higher vitality than other green spaces located in urbanized areas. In contrast, schools with a low green space ratio in their school zone are mostly situated in high-density residential areas and the green spaces show relatively low vegetation vitality. Except for the urban forest, the majority of urban green spaces in urbanized areas are landscape green facilities in apartment districts. The other types of urban open spaces such as environmentally shaped schoolyards or street parks account only for a very small proportion of school surroundings. Therefore, it is needed to establish countermeasures in the context of urban planning; e.g. to promote the school forest projects preferentially by selecting schools with a extremely low green space ratio in their school zone, to foster roof greening in near surroundings, and to connect schoolyards organically with nearby apartment landscape green facilities as an easily accessible urban open space.

• Korean Journal of Ecology and Environment
• /
• v.51 no.1
• /
• pp.105-123
• /
• 2018

Currently, satellite images act as essential and important data in water resources, environment, and ecology as well as information of geographic information system. In this paper, we will investigate basic characteristics of satellite images, especially application examples in water resources. In recent years, researches on spatial and temporal characteristics of large-scale regions utilizing the advantages of satellite imagery have been actively conducted for fundamental hydrological components such as evapotranspiration, soil moisture and natural disasters such as drought, flood, and heavy snow. Furthermore, it is possible to analyze temporal and spatial characteristics such as vegetation characteristics, plant production, net primary production, turbidity of water bodies, chlorophyll concentration, and water quality by using various image information utilizing various sensor information of satellites. Korea is planning to launch a satellite for water resources and environment in the near future, so various researches are expected to be activated on this field.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.3 no.4
• /
• pp.199-205
• /
• 2001

This study was conducted to figure out temperature profiles of a partially developed paddy rice canopy, which are necessary to run plant disease forecasting models. Air temperature over and within the developing rice canopy was monitored from one month after transplanting (June 29) to just before heading (August 24) in 1999 and 2001. During the study period, the temporal march of the within-canopy profile was analyzed and an empirical formula was developed for simulating the profile. A partially developed rice canopy temperature seemed to be controlled mainly by the ambient temperature above the canopy and the water temperature beneath the canopy, and to some extent by the solar altitude, resulting in alternating isothermal and inversion structures. On sunny days, air temperature at the height of maximum leafages was increased at the same rate as the ambient temperature above the canopy after sunrise. Below the height, the temperature increase was delayed until the solar noon. Air temperature near the water surface varied much less than those of the outer- and the upper-canopy, which kept increasing by the time of daily maximum temperature observed at the nearby synoptic station. After sunset, cooling rate is much less at the lower canopy, resulting in an isothermal profile at around the midnight. A fairly consistent drop in temperature at rice paddies compared with the nearby synoptic weather stations across geographic areas and time of day was found. According to this result, a cooling by 0.6 to 1.2$^{\circ}C$ is expected over paddy rice fields compared with the officially reported temperature during the summer months. An empirical equation for simulating the temperature profile was formulated from the field observations. Given the temperature estimates at 150 cm above the canopy and the maximum deviation at the lowest layer, air temperature at any height within the canopy can be predicted by this equation. As an application, temperature surfaces at several heights within rice fields were produced over the southwestern plains in Korea at a 1 km by 1km grid spacing, where rice paddies were identified by a satellite image analysis. The outer canopy temperature was prepared by a lapse rate corrected spatial interpolation of the synoptic temperature observations combined with the hourly cooling rate over the rice paddies.