• Korean Journal of Remote Sensing
• /
• v.40 no.1
• /
• pp.57-69
• /
• 2024

Crop classification plays a vitalrole in monitoring agricultural landscapes and enhancing food production. In this study, we explore the effectiveness of Long Short-Term Memory (LSTM) models for crop classification, focusing on distinguishing between apple and rice crops. The aim wasto overcome the challenges associatedwith finding phenology-based classification thresholds by utilizing LSTM to capture the entire Normalized Difference Vegetation Index (NDVI)trend. Our methodology involvestraining the LSTM model using a reference site and applying it to three separate three test sites. Firstly, we generated 25 NDVI imagesfrom the Sentinel-2A data. Aftersegmenting study areas, we calculated the mean NDVI values for each segment. For the reference area, employed a training approach utilizing the NDVI trend line. This trend line served as the basis for training our crop classification model. Following the training phase, we applied the trained model to three separate test sites. The results demonstrated a high overall accuracy of 0.92 and a kappa coefficient of 0.85 for the reference site. The overall accuracies for the test sites were also favorable, ranging from 0.88 to 0.92, indicating successful classification outcomes. We also found that certain phenological metrics can be less effective in crop classification therefore limitations of relying solely on phenological map thresholds and emphasizes the challenges in detecting phenology in real-time, particularly in the early stages of crops. Our study demonstrates the potential of LSTM models in crop classification tasks, showcasing their ability to capture temporal dependencies and analyze timeseriesremote sensing data.While limitations exist in capturing specific phenological events, the integration of alternative approaches holds promise for enhancing classification accuracy. By leveraging advanced techniques and considering the specific challenges of agricultural landscapes, we can continue to refine crop classification models and support agricultural management practices.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_3
• /
• pp.891-907
• /
• 2023

Compact Advanced Satellite 500-4 (CAS500-4) is scheduled to be launched to collect high spatial resolution data focusing on vegetation applications. To achieve this goal, accurate surface reflectance retrieval through atmospheric correction is crucial. Therefore, a machine learning-based atmospheric correction algorithm was developed to simulate atmospheric correction from a radiative transfer model using Sentinel-2 data that have similarspectral characteristics as CAS500-4. The algorithm was then evaluated mainly for forest areas. Utilizing the atmospheric correction parameters extracted from Sentinel-2 and GEOKOMPSAT-2A (GK-2A), the atmospheric correction algorithm was developed based on Random Forest and Light Gradient Boosting Machine (LGBM). Between the two machine learning techniques, LGBM performed better when considering both accuracy and efficiency. Except for one station, the results had a correlation coefficient of more than 0.91 and well-reflected temporal variations of the Normalized Difference Vegetation Index (i.e., vegetation phenology). GK-2A provides Aerosol Optical Depth (AOD) and water vapor, which are essential parameters for atmospheric correction, but additional processing should be required in the future to mitigate the problem caused by their many missing values. This study provided the basis for the atmospheric correction of CAS500-4 by developing a machine learning-based atmospheric correction simulation algorithm.

• Geomechanics and Engineering
• /
• v.37 no.4
• /
• pp.307-321
• /
• 2024

This paper delves into the critical assessment of predicting sidewall displacement in underground caverns through the application of nine distinct machine learning techniques. The accurate prediction of sidewall displacement is essential for ensuring the structural safety and stability of underground caverns, which are prone to various geological challenges. The dataset utilized in this study comprises a total of 310 data points, each containing 13 relevant parameters extracted from 10 underground cavern projects located in Iran and other regions. To facilitate a comprehensive evaluation, the dataset is evenly divided into training and testing subset. The study employs a diverse array of machine learning models, including recurrent neural network, back-propagation neural network, K-nearest neighbors, normalized and ordinary radial basis function, support vector machine, weight estimation, feed-forward stepwise regression, and fuzzy inference system. These models are leveraged to develop predictive models that can accurately forecast sidewall displacement in underground caverns. The training phase involves utilizing 80% of the dataset (248 data points) to train the models, while the remaining 20% (62 data points) are used for testing and validation purposes. The findings of the study highlight the back-propagation neural network (BPNN) model as the most effective in providing accurate predictions. The BPNN model demonstrates a remarkably high correlation coefficient (R2 = 0.99) and a low error rate (RMSE = 4.27E-05), indicating its superior performance in predicting sidewall displacement in underground caverns. This research contributes valuable insights into the application of machine learning techniques for enhancing the safety and stability of underground structures.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.26 no.2
• /
• pp.115-125
• /
• 2024

Wind is a meteorological factor that has a significant impact on agriculture. Gust cause damage such as fruit drop and damage to facilities. In this study, low-altitude wind speed prediction was performed by applying physical models to Local Ensemble Prediction System (LENS). Logarithmic Law (LOG) and Power Law (POW) were used as the physical models, and Korea Ministry of Environment indicators and Moderate Resolution Imaging Spectroradiometer (MODIS) data were applied as indicator variables. We collected and verified wind and gust data at 3m altitude in 2022 operated by the Rural Development Administration, and presented the results in scatter plot, correlation coefficient, Root Mean Square Error (RMSE), Normalized Root Mean Square Error (NRMSE), and Threat Score (TS). The LOG-applied model showed better results in wind speed, and the POW-applied model showed better results in gust.

• Korean Journal of Remote Sensing
• /
• v.39 no.6_1
• /
• pp.1341-1352
• /
• 2023

The normalized difference vegetation index (NDVI) derived from satellite images is a crucial tool to monitor forests and agriculture for broad areas because the periodic acquisition of the data is ensured. However, optical sensor-based vegetation indices(VI) are not accessible in some areas covered by clouds. This paper presented a synthetic aperture radar (SAR) based approach to retrieval of the optical sensor-based NDVI using machine learning. SAR system can observe the land surface day and night in all weather conditions. Radar vegetation indices (RVI) from the Sentinel-1 vertical-vertical (VV) and vertical-horizontal (VH) polarizations, surface elevation, and air temperature are used as the input features for an automated machine learning (AutoML) model to conduct the gap-filling of the Sentinel-2 NDVI. The mean bias error (MAE) was 7.214E-05, and the correlation coefficient (CC) was 0.878, demonstrating the feasibility of the proposed method. This approach can be applied to gap-free nationwide NDVI construction using Sentinel-1 and Sentinel-2 images for environmental monitoring and resource management.

• Journal of the Korean earth science society
• /
• v.36 no.1
• /
• pp.109-124
• /
• 2015

A novel approach, hybrid surface rainfall (KNU-HSR) technique developed by Kyungpook Natinal University, was utilized for improving the radar rainfall estimation. The KNU-HSR technique estimates radar rainfall at a 2D hybrid surface consistings of the lowest radar bins that is immune to ground clutter contaminations and significant beam blockage. Two HSR techniques, static and dynamic HSRs, were compared and evaluated in this study. Static HSR technique utilizes beam blockage map and ground clutter map to yield the hybrid surface whereas dynamic HSR technique additionally applies quality index map that are derived from the fuzzy logic algorithm for a quality control in real time. The performances of two HSRs were evaluated by correlation coefficient (CORR), total ratio (RATIO), mean bias (BIAS), normalized standard deviation (NSD), and mean relative error (MRE) for ten rain cases. Dynamic HSR (CORR=0.88, BIAS= $-0.24mm\;hr^{-1}$, NSD=0.41, MRE=37.6%) shows better performances than static HSR without correction of reflectivity calibration bias (CORR=0.87, BIAS= $-2.94mm\;hr^{-1}$, NSD=0.76, MRE=58.4%) for all skill scores. Dynamic HSR technique overestimates surface rainfall at near range whereas it underestimates rainfall at far ranges due to the effects of beam broadening and increasing the radar beam height. In terms of NSD and MRE, dynamic HSR shows the best results regardless of the distance from radar. Static HSR significantly overestimates a surface rainfall at weaker rainfall intensity. However, RATIO of dynamic HSR remains almost 1.0 for all ranges of rainfall intensity. After correcting system bias of reflectivity, NSD and MRE of dynamic HSR are improved by about 20 and 15%, respectively.

• Investigative Magnetic Resonance Imaging
• /
• v.18 no.2
• /
• pp.120-132
• /
• 2014

Purpose : To compare dynamic susceptibility contrast imaging, diffusion-weighted imaging, and susceptibility-weighted imaging (SWI) for the differentiation of tumor recurrence and delayed radiation therapy (RT)-related changes in patients treated with RT for primary brain tumors. Materials and Methods: We enrolled 24 patients treated with RT for various primary brain tumors, who showed newly appearing enhancing lesions more than one year after completion of RT on follow-up MRI. The enhancing-lesions were confirmed as recurrences (n=14) or RT-changes (n=10). We calculated the mean values of normalized cerebral blood volume (nCBV), apparent diffusion coefficient (ADC), and proportion of dark signal intensity on SWI (proSWI) for the enhancing-lesions. All the values between the two groups were compared using t-test. A multivariable logistic regression model was used to determine the best predictor of differential diagnosis. The cutoff value of the best predictor obtained from receiver-operating characteristic curve analysis was applied to calculate the sensitivity, specificity, and accuracy for the diagnosis. Results: The mean nCBV value was significantly higher in the recurrence group than in the RT-change group (P=.004), and the mean proSWI was significantly lower in the recurrence group (P<.001). However, no significant difference was observed in the mean ADC values between the two groups. A multivariable logistic regression analysis showed that proSWI was the only independent variable for the differentiation; the sensitivity, specificity, and accuracy were 78.6% (11 of 14), 100% (10 of 10), and 87.5% (21 of 24), respectively. Conclusion: The proSWI was the most promising parameter for the differentiation of newly developed enhancing-lesions more than one year after RT completion in brain tumor patients.

• Journal of Environmental Impact Assessment
• /
• v.23 no.6
• /
• pp.432-444
• /
• 2014

When the water quality modelling is done with a manual calibration, it is possible that the researcher's opinion may affect the objectivity of the research. Hence, the role of the automatic calibration is highly important. This research applies a technique to automatically calibrate the water quality parameters by implementing an optimization method. This involves estimating the optimum water quality parameters targeting influential parameters towards the lake's BOD, DO, Phosphorus, Nitrogen and Phytoplankton. To accurately calculate the water temperature and hydraulic characteristics of a deep, stratifying lake, EFDC, a 3-dimensional hydraulic model which can be linked to the WASP7 was applied. With EFDC, the segment of the lake is formed and utilized as an input data of the WASP7. For the calibration of the water quality parameters of the WASP7, an influence coefficient algorithm and a genetic algorithm was applied. Of the five water quality variables for calibration, the normalized residuals of the observed and calculated values of DO, TN, CBOD were relatively small and the three water quality variables were calibrated properly. Yet the accuracy of the calibration of TP and Chl-a was relatively low.

• Korean Journal of Soil Science and Fertilizer
• /
• v.36 no.4
• /
• pp.200-209
• /
• 2003

Distribution coefficient ($K_d$) is an universal parameter estimating cadmium partition for a soil-water-crop system in agricultural lands. This study was performed to find some factors affecting soil-water partition coefficients for cadmium in some Korean soils. The distribution coefficients ($K_d$) of cadmium for the 15 series of agricultural soils were measured at quasi-steady state in the pH ranges from 2 to 11. The adsorption data of the selected soils showed a linear relationship between log $K_d$ and pH, which was well agreed with theoretically expected results ; $log\;K_d=0.6339pH+0.5532(r^2=0.70^{**})$. Normalization of the partition coefficients were performed in a range of pH 3.5 ~ 8.5 to minimize adverse effects of Al dissolution, cationic competition, and organic matter dissolution. The $K_d$-om, partition coefficients normalized for organic matter, improved this linearity to the pH of soils. The values of $K_d$-om measured from the field samples were significantly correlated with those of $K_d$ predicted from the sorption-edge experimental data ($r^2=0.68^{**}$).

• Journal of Biosystems Engineering
• /
• v.43 no.2
• /
• pp.148-159
• /
• 2018

Purpose: Unmanned air vehicle (UAV) remote sensing was applied to test various vegetation indices and make prediction models of protein content of rice for monitoring grain quality and proper management practice. Methods: Image acquisition was carried out by using NIR (Green, Red, NIR), RGB and RE (Blue, Green, Red-edge) camera mounted on UAV. Sampling was done synchronously at the geo-referenced points and GPS locations were recorded. Paddy samples were air-dried to 15% moisture content, and then dehulled and milled to 92% milling yield and measured the protein content by near-infrared spectroscopy. Results: Artificial neural network showed the better performance with $R^2$ (coefficient of determination) of 0.740, NSE (Nash-Sutcliffe model efficiency coefficient) of 0.733 and RMSE (root mean square error) of 0.187% considering all 54 samples than the models developed by PR (polynomial regression), SLR (simple linear regression), and PLSR (partial least square regression). PLSR calibration models showed almost similar result with PR as 0.663 ($R^2$) and 0.169% (RMSE) for cloud-free samples and 0.491 ($R^2$) and 0.217% (RMSE) for cloud-shadowed samples. However, the validation models performed poorly. This study revealed that there is a highly significant correlation between NDVI (normalized difference vegetation index) and protein content in rice. For the cloud-free samples, the SLR models showed $R^2=0.553$ and RMSE = 0.210%, and for cloud-shadowed samples showed 0.479 as $R^2$ and 0.225% as RMSE respectively. Conclusion: There is a significant correlation between spectral bands and grain protein content. Artificial neural networks have the strong advantages to fit the nonlinear problem when a sigmoid activation function is used in the hidden layer. Quantitatively, the neural network model obtained a higher precision result with a mean absolute relative error (MARE) of 2.18% and root mean square error (RMSE) of 0.187%.