• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.16 no.1
• /
• pp.26-37
• /
• 2017

The study is to develop safety performance functions(SPFs) for urban intersections using random effects Tobit regression model and to analyze correlations between crashes and factors. Also fixed effects Tobit regression model was estimated to compare and analyze model validation with random effects model. As a result, AADT, speed limits, number of lanes, land usage, exclusive right turn lanes and front traffic signal were found to be significant. For comparing statistical significance between random and fixed effects model, random effects Tobit regression model of total crash rate could be better statistical significance with $R^2_p$ : 0.418, log-likelihood at convergence: -3210.103, ${\rho}^2$: 0.056, MAD: 19.533, MAPE: 75.725, RMSE: 26.886 comparing with $R^2_p$ : 0.298, log-likelihood at convergence: -3276.138, ${\rho}^2$: 0.037, MAD: 20.725, MAPE: 82.473, RMSE: 27.267 for the fixed model. Also random effects Tobit regression model of injury crash rate has similar results of model statistical significant with random effects Tobit regression model.

• Korean Journal of Remote Sensing
• /
• v.36 no.2_2
• /
• pp.309-323
• /
• 2020

Estimation of chlorophyll a concentration (CHL) on coastal areas using remote sensing has been mostly performed through multi-spectral satellite image analysis. Recently, various studies using hyperspectral imagery have been attempted. In particular, airborne hyperspectral imagery is composed of hundreds of bands with a narrow band width and high spatial resolution, and thus may be more effective in coastal areas than estimation of CHL through conventional satellite image. In this study, comparative analysis of hyperspectral and satellite-based CHL images was performed to estimate CHL in coastal areas. As a result of analyzing CHL and seawater spectrum data obtained by field survey conducted on the south coast of Korea, the seawater spectrum with high CHL peaked near the wavelength bands of 570 and 680 nm. Using this spectral feature, a new band ratio of 570 / 490 nm for estimating CHL was proposed. Through regression analysis between band ratio and the measured CHL were generated new CHL empirical formula. Validation of new empirical formula using the measured CHL showed valid results, with R² of 0.70, RMSE of 2.43 mg m^-3, and mean bias of 3.46 mg m^-3. As a result of applying the new empirical formula to hyperspectral and satellite images, the average RMSE between hyperspectral imagery and the measured CHL was 0.12 mg m^-3, making it possible to estimate CHL with higher accuracy than multi-spectral satellite images. Through these results, it is expected that it is possible to provide more accurate and precise spatial distribution information of CHL in coastal areas by utilizing hyperspectral imagery.

• Korean Journal of Remote Sensing
• /
• v.37 no.6_1
• /
• pp.1669-1683
• /
• 2021

AI intelligent agriculture and digital agriculture are important for the science of agriculture. Leaf chlorophyll contents(LCC) are one of the most important indicators to determine the growth status of vegetable crops. In this study, a support vector machine (SVM) regression model was produced using an unmanned aerial vehicle-based RGB camera and a multispectral (MSP) sensor for onions and garlic, and the LCC estimation applicability of the RGB camera was reviewed by comparing it with the MSP sensor. As a result of this study, the RGB-based LCC model showed lower results than the MSP-based LCC model with an average R² of 0.09, RMSE 18.66, and nRMSE 3.46%. However, the difference in accuracy between the two sensors was not large, and the accuracy did not drop significantly when compared with previous studies using various sensors and algorithms. In addition, the RGB-based LCC model reflects the field LCC trend well when compared with the actual measured value, but it tends to be underestimated at high chlorophyll concentrations. It was possible to confirm the applicability of the LCC estimation with RGB considering the economic feasibility and versatility of the RGB camera. The results obtained from this study are expected to be usefully utilized in digital agriculture as AI intelligent agriculture technology that applies artificial intelligence and big data convergence technology.

• KIPS Transactions on Computer and Communication Systems
• /
• v.10 no.3
• /
• pp.71-80
• /
• 2021

The selection of an appropriate neural network algorithm is an important step for accurate data prediction in machine learning. Many algorithms based on basic artificial neural networks have been devised to efficiently predict future data. These networks include deep neural networks (DNNs), recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and gated recurrent unit (GRU) neural networks. Developers face difficulties when choosing among these networks because sufficient information on their performance is unavailable. To alleviate this difficulty, we evaluated the performance of each algorithm by comparing their errors and processing times. Each neural network model was trained using a tax dataset, and the trained model was used for data prediction to compare accuracies among the various algorithms. Furthermore, the effects of activation functions and various optimizers on the performance of the models were analyzed The experimental results show that the GRU and LSTM algorithms yields the lowest prediction error with an average RMSE of 0.12 and an average R2 score of 0.78 and 0.75 respectively, and the basic DNN model achieves the lowest processing time but highest average RMSE of 0.163. Furthermore, the Adam optimizer yields the best performance (with DNN, GRU, and LSTM) in terms of error and the worst performance in terms of processing time. The findings of this study are thus expected to be useful for scientists and developers.

• Korean Journal of Remote Sensing
• /
• v.28 no.1
• /
• pp.69-77
• /
• 2012

Surface insolation is one of the major indicators for climate research over the Earth system. For the climate research, long-term data and wide range of spatial coverage from the data observed by two or more of satellites of the same orbit are needed. It is important to improve the continuity and consistency of the derived products, such as surface insolation, from different satellites. In this study, surface insolations based on Geostationary Operational Environmental Satellite (GOES-9) and Multi-functional Transport Satellites (MTSAT-1R) were compared during overlap period using physical model of insolation to find ways to improve the consistency and continuity between two satellites through comparison of each channel data and ground observation data. The thermal infrared brightness temperature of two satellites show a relatively good agreement between two satellites : rootmean square error (RMSE)=5.595 Kelvin; Bias=2.065 Kelvin. Whereas, visible channels shown a quite different values, but it distributed similar tendency. And the surface insolations from two satellites are different from the ground observation data. To improve the quality of retrieved insolations, we have reproduced surface insolation of each satellite through adjustment of the Cloud Factor, and the Cloud Factor for GOES-9 satellite is modified based on the analysis result of difference channel data. As a result, the insolations estimated from GOES-9 for cloudy conditions show good agreement with MTSAT-1R and ground observation : RMSE=$83.439W\;m^{-2}$ Bias=$27.296W\;m^{-2}$. The result improved accuracy confirms that the modification of Cloud Factor for GOES-9 can improve the continuity and consistency of the insolations derived from two or more satellites.

• 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.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.31 no.5B
• /
• pp.393-403
• /
• 2011

This study theoretically reviewed the empirical G/R ratio by considering three regression and trend lines; the general linear regression curve, linear regression curve passing the origin, and the line passing the origin and the mass center of observed data. This review included the problem of choosing the independent variable and that of considering the zero measurements. This review result was also applied to the Typhoon Maemi in 2003 for their evaluation. Additionally, those regression and trend lines were compared using the RMSE between the corrected radar rainfall and observed rain gauge rainfall to select the most appropriate G/R ratio. Summarizing the results is as follows. First, the results of selecting the rain gauge rainfall as the independent variable were found better than the opposite case. Second, the effect of zero measurements varies depending on the structure of radar and rain gauge rainfall. Finally, the results from the comparison of three regression and trend lines shows that the slope of the regression line passing the origin with its independent variable of rain gauge rainfall would be used most appropriately for the G/R ratio, especially when the corrected radar rainfall is used for the flood analysis. The effect of zero measurements in this case was found not so significant.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.4
• /
• pp.487-494
• /
• 2023

In this study, performance comparison was performed on R and TensorFlow, which are free deep learning tools. In the experiment, six types of deep neural networks were built using each tool, and the neural networks were trained using the 10-year Korean temperature dataset. The number of nodes in the input layer of the constructed neural network was set to 10, the number of output layers was set to 5, and the hidden layer was set to 5, 10, and 20 to conduct experiments. The dataset includes 3600 temperature data collected from Gangnam-gu, Seoul from March 1, 2013 to March 29, 2023. For performance comparison, the future temperature was predicted for 5 days using the trained neural network, and the root mean square error (RMSE) value was measured using the predicted value and the actual value. Experiment results shows that when there was one hidden layer, the learning error of R was 0.04731176, and TensorFlow was measured at 0.06677193, and when there were two hidden layers, R was measured at 0.04782134 and TensorFlow was measured at 0.05799060. Overall, R was measured to have better performance. We tried to solve the difficulties in tool selection by providing quantitative performance information on the two tools to users who are new to machine learning.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.61 no.5
• /
• pp.69-77
• /
• 2019

The objective of this study was to calibrate and validate the HSPF (Hydrological Simulation Program-Fortran) model for estimating the runoff of the Hapcheon dam watershed. Spatial data, such as watershed, stream, land use, and a digital elevation map, were used as input data for the HSPF model. Observed runoff data from 2000 to 2016 in study watershed were used for calibration and validation. Hydrologic parameters for runoff calibration were selected based on the user's manual and references, and trial and error method was used for parameter calibration. The $R^2$, RMSE (root-mean-square error), RMAE (relative mean absolute error), and NSE (Nash-Sutcliffe efficiency coefficient) were used to evaluate the model's performance. Calibration and validation results showed that annual mean runoff was within ${\pm}4%$ error. The model performance criteria for calibration and validation showed that $R^2$ was in the rang of 0.78 to 0.83, RMSE was 2.55 to 2.76 mm/day, RMAE was 0.46 to 0.48 mm/day, and NSE was 0.81 to 0.82 for daily runoff. The amount of inflow to Hapcheon Dam was calculated from the calibrated HSPF model and the result was compared with observed inflow, which was -0.9% error. As a result of analyzing the relation between inflow and storage capacity, it was found that as the inflow increases, the storage increases, and when the inflow decreases, the storage also decreases. As a result of correlation between inflow and storage, $R^2$ of the measured inflow and storage was 0.67, and the simulated inflow and storage was 0.61.

• Journal of Oral Medicine and Pain
• /
• v.46 no.2
• /
• pp.21-32
• /
• 2021

Purpose: The purpose of the present study was to investigate the correlation between age and 34 radiomorphometric parameters on panoramic radiographs, and to provide generalized linear models (GLMs) as a non-invasive, inexpensive, and accurate method to the forensic judgement of living individual's age. Methods: The study included 417 digital panoramic radiographs of Korean individuals (178 males and 239 females, mean age: 32.57±17.81 years). Considering the skeletal differences between the sexes, GLMs were obtained separately according to sex, as well as across the total sample. For statistical analysis and to predict the accuracy of the new GLMs, root mean squared error (RMSE) and adjusted R-squared (R²) were calculated. Results: The adjusted R²-values of the developed GLMs in the total sample, and male and female groups were 0.623, 0.637, and 0.660, respectively (p<0.001), while the allowable RMSE values were 8.80, 8.42, and 8.53 years, respectively. In the GLM of the total sample, the most influential predictor of greater age was decreased pulp area in the #36 first molar (beta=-26.52; p<0.01), followed by the presence of periodontitis (beta=10.24; p<0.01). In males, the most influential factor was the presence of periodontitis (beta=9.20; p<0.05), followed by the number of full veneer crowns (beta=2.19; p<0.001). In females, the most influential predictor was the presence of periodontitis (beta=18.10; p<0.001), followed by the tooth area of the #16 first molar (beta=-11.57; p<0.001). Conclusions: We established acceptable GLM for each sex and found out the predictors necessary to age estimation which can be easily found in panoramic radiographs. Our study provides reference that parameters such as the area of tooth and pulp, the number of teeth treated, and the presence of periodontitis should be considered in estimating age.