• Journal of Biosystems Engineering
• /
• v.33 no.5
• /
• pp.362-369
• /
• 2008

There is developing interest in precision agriculture in Korea, despite the fact that typical Korean fields are less than 1 ha in size. Describing within-field variability in typical Korean production settings is a fundamental first step toward determining the size of management zones and the inter-relationships between limiting factors, for establishment of site-specific management strategies. Measurements of rice (Oriza Sativa L) yield, chlorophyll content, and soil properties were obtained in a small (100-m by 30-m) Korean rice paddy field. Yield data were manually collected on 10-m by 5-m grids (180 samples with 3 samples in each of 60 grid cells) and chlorophyll content was measured using a Minolta SPAD 502 in 2-m by 2-m grids. Soil samples were collected at 275 points to compare results from sampling at different scales. Ten soil properties important for rice production in Korea were determined through laboratory analyses. Variogram analysis and point kriging with and without median polishing were conducted to determine the variability of the measured parameters. Influence of variogram model selection and other parameters on the interpretation of the data was investigated. For many of the data, maximum values were greater than double the minimum values, indicating considerable spatial variability in the small paddy field, and large-scale spatial trends were present. When variograms were fit to the original data, the limits of spatial dependency for rice yield and SP AD reading were 11.5 m and 6.5 m, respectively, and after detrending the limits were reduced to 7.4 m and 3.9 m. The range of spatial dependency for soil properties was variable, with several having ranges as short as 2 m and others having ranges greater than 30 m. Kriged maps of the variables clearly showed the presence of both large-scale (trend) variability and small-scale variability in this small field where it would be reasonable to expect uniformity. These findings indicate the potential for applying the principles and technology of precision agriculture for Korean paddy fields. Additional research is needed to confirm the results with data from other fields and crops.d similar tendency with the result for the frequency less than 20 Hz, but the width of change was reduced highly.

• Journal of Korea Water Resources Association
• /
• v.51 no.11
• /
• pp.1031-1041
• /
• 2018

ANFIS-based methodology for improving GloSea5 ensemble weather forecast is developed and evaluated in this study. The proposed method consists of two steps: pre & post processing. For ensemble prediction of GloSea5, weights are assigned to the ensemble members based on Optimal Weighting Method (OWM) in the pre-processing. Then, the bias of the results of pre-processed is corrected based on Model Output Statistics (MOS) method in the post-processing. The watershed of the Chungju multi-purpose dam in South Korea is selected as a study area. The results of evaluation indicated that the pre-processing step (CASE1), the post-processing step (CASE2), pre & post processing step (CASE3) results were significantly improved than the original GloSea5 bias correction (BC_GS5). Correction performance is better the order of CASE3, CASE1, CASE2. Also, the accuracy of pre-processing was improved during the season with high variability of precipitation. The post-processing step reduced the error that could not be smoothed by pre-processing step. It could be concluded that this methodology improved the ability of GloSea5 ensemble weather forecast by using ANFIS, especially, for the summer season with high variability of precipitation when applied both pre- and post-processing steps.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.27 no.8
• /
• pp.1426-1435
• /
• 2003

Design goal is to find the one that has the highest probability of success and the smallest variation. A robustness index has been proposed to satisfy these conditions. The two-step optimization process of the target problem requires a scaling factor. The search process of a scaling factor is replaced with the making of the decoupled design between the mean and the standard deviation. The decoupled design matrix is formed from the sensitivity or the sum of squares. After establishing the design matrix, the robust design process has a new three-step one. The first is ″reduce variability,″ the second is ″make the candidate designs that satisfy constraints and move the mean on the target,″ and the final is ″select the best robust design using the proposed robustness index.″ The robust design process is verified by three examples and the results using the robustness index are compared with those of other indices.

• International Journal of Reliability and Applications
• /
• v.9 no.1
• /
• pp.17-30
• /
• 2008

Analysis of measurement systems is important to determine if the measurement process is adequate to measure the part-to-part variability in the process. Control chart techniques provide an effective, and easy-to-use method for performing this analysis. However, application with the real data for the evaluation procedure for multiple measurement systems have not been demonstrated. This research will provide a methodology for the evaluation of part-to-part variation and variation of different measurement systems step by step followed by number of case studies for each methodologies provided.

• Atmosphere
• /
• v.26 no.4
• /
• pp.541-551
• /
• 2016

A real-time quality control algorithm for $PM_{10}$ concentration measured by Continuous Ambient Particulate Monitor (FH62C14, Thermo Fisher Scientific Inc.) has been developed. The quality control algorithm for $PM_{10}$ data consists of five main procedures. The first step is valid value check. The values should be within the acceptable range limit. Upper ($5,000{\mu}g\;m^{-3}$) and lower ($0{\mu}g\;m^{-3}$) values of instrument detectable limit have to be eliminated as being unrealistic. The second step is valid error check. Whenever unusual condition occurs, the instrument will save error code. Value having an error code is eliminated. The third step is persistence check. This step checks on a minimum required variability of data during a certain period. If the $PM_{10}$ data do not vary over the past 60 minutes by more than the specific limit ($0{\mu}g\;m^{-3}$) then the current 5-minute value fails the check. The fourth step is time continuity check, which is checked to eliminate gross outlier. The last step is spike check. The spikes in the time series are checked. The outlier detection is based on the double-difference time series, using the median. Flags indicating normal and abnormal are added to the raw data after quality control procedure. The quality control algorithm is applied to $PM_{10}$ data for Asian dust and non-Asian dust case at Seoul site and dataset for the period 2013~2014 at 26 sites in Korea.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.8A
• /
• pp.639-647
• /
• 2012

Vision-based on-road vehicle detection is one of the key techniques in automotive driver assistance systems. However, due to the huge within-class variability in vehicle appearance and environmental changes, it remains a challenging task to develop an accurate and reliable detection system. In general, a vehicle detection system consists of two steps. The candidate locations of vehicles are found in the Hypothesis Generation (HG) step, and the detected locations in the HG step are verified in the Hypothesis Verification (HV) step. Since the final decision is made in the HV step, the HV step is crucial for accurate detection. In this paper, we propose using a reduced multivariate polynomial pattern classifier (RM) for the HV step. Our experimental results show that the RM classifier outperforms the well-known Support Vector Machine (SVM) classifier, particularly in terms of the fast decision speed, which is suitable for real-time implementation.

• PNF and Movement
• /
• v.15 no.3
• /
• pp.361-371
• /
• 2017

Purpose: The purpose of this study was to examine the effects of performing a dual task on gait velocity, temporospatial variables, and symmetry in subjects with subacute stroke. Methods: The study included 14 independent community ambulators with gait velocity of 0.8m/s. The Korean mini-mental state examination, the Berg balance scale, the Trunk impairment scale, and the Fugl-Meyer assessment scale were used to recruit homogeneous subjects. Subjects performed a single task (10m ambulation at a comfortable speed) and a dual task (10m ambulation at a comfortable speed while carrying a water-filled glass). Gait variables were examined with the OptoGait system. Results: The findings of this study were as follows: 1) Gait velocity decreased significantly in the dual-task condition as compared to the single task condition. 2) There were no significant differences between the paretic and non-paretic stances. 3) Paretic swing decreased significantly in the dual-task condition as compared to the single task condition. 4) The non-paretic, double-limb support phase increased significantly in the dual-task condition as compared to the single- task condition. 5) There was no significant difference in temporal symmetry. 6) Non-paretic step length decreased significantly in the dual-task condition as compared to the single-task condition. 7) There was no significant difference in spatial symmetry. Conclusion: Performing dual tasks decreases gait velocity, paretic swing phase, and non-paretic step length, while it increases non-paretic double limb support. In addition, although there is no difference in temporospatial symmetry, there is high inter-subject variability in temporospatial symmetry. Thus, dual tasks should be selected in accordance with the functional level of the hemiplegic patient, and inter-subject variability of the individual should be considered when dual tasks are considered for gait-training of hemiplegic patients.

• Journal of Korean Medical Science
• /
• v.33 no.43
• /
• pp.239.1-239.12
• /
• 2018

Background: We described a novel multi-step retinal fundus image reading system for providing high-quality large data for machine learning algorithms, and assessed the grader variability in the large-scale dataset generated with this system. Methods: A 5-step retinal fundus image reading tool was developed that rates image quality, presence of abnormality, findings with location information, diagnoses, and clinical significance. Each image was evaluated by 3 different graders. Agreements among graders for each decision were evaluated. Results: The 234,242 readings of 79,458 images were collected from 55 licensed ophthalmologists during 6 months. The 34,364 images were graded as abnormal by at-least one rater. Of these, all three raters agreed in 46.6% in abnormality, while 69.9% of the images were rated as abnormal by two or more raters. Agreement rate of at-least two raters on a certain finding was 26.7%-65.2%, and complete agreement rate of all-three raters was 5.7%-43.3%. As for diagnoses, agreement of at-least two raters was 35.6%-65.6%, and complete agreement rate was 11.0%-40.0%. Agreement of findings and diagnoses were higher when restricted to images with prior complete agreement on abnormality. Retinal/glaucoma specialists showed higher agreements on findings and diagnoses of their corresponding subspecialties. Conclusion: This novel reading tool for retinal fundus images generated a large-scale dataset with high level of information, which can be utilized in future development of machine learning-based algorithms for automated identification of abnormal conditions and clinical decision supporting system. These results emphasize the importance of addressing grader variability in algorithm developments.

• Imaging Science in Dentistry
• /
• v.51 no.3
• /
• pp.299-306
• /
• 2021

Purpose: This study aimed to propose a fully automatic landmark identification model based on a deep learning algorithm using real clinical data and to verify its accuracy considering inter-examiner variability. Materials and Methods: In total, 950 lateral cephalometric images from Yonsei Dental Hospital were used. Two calibrated examiners manually identified the 13 most important landmarks to set as references. The proposed deep learning model has a 2-step structure-a region of interest machine and a detection machine-each consisting of 8 convolution layers, 5 pooling layers, and 2 fully connected layers. The distance errors of detection between 2 examiners were used as a clinically acceptable range for performance evaluation. Results: The 13 landmarks were automatically detected using the proposed model. Inter-examiner agreement for all landmarks indicated excellent reliability based on the 95% confidence interval. The average clinically acceptable range for all 13 landmarks was 1.24 mm. The mean radial error between the reference values assigned by 1 expert and the proposed model was 1.84 mm, exhibiting a successful detection rate of 36.1%. The A-point, the incisal tip of the maxillary and mandibular incisors, and ANS showed lower mean radial error than the calibrated expert variability. Conclusion: This experiment demonstrated that the proposed deep learning model can perform fully automatic identification of cephalometric landmarks and achieve better results than examiners for some landmarks. It is meaningful to consider between-examiner variability for clinical applicability when evaluating the performance of deep learning methods in cephalometric landmark identification.

• Toxicological Research
• /
• v.6 no.2
• /
• pp.205-213
• /
• 1990

The regulation of neurotoxicants has usually been based upon setting reference doses by dividing a no observed adverse effect level (NOAEL) by uncertainty factors that theoretically account for interspecies and intraspecies extraploation of experimental results in animals to humans. Recently, we have proposed a four-step alternative procedure which provides quantitative estimates of risk as a function of dose. The first step is to establish a mathematical relationship between a biological effect or biomarker and the dose of chemical administered. The second step is to determine the distribution (variability) of individual measurements of biological effects or their biomarkers about the dose response curve. The third step is to define an adverse or abnormal level of a biological effect or biomarker in an untreated population. The fourth and final step is to combine the information from the first three steps to estimate the risk (proportion of individuals exceeding on adverse or abnormal level of a biological effect or biomarker) as a function of dose. The primary purpose of this report is to enhance the certainty of the first step of this procedure by improving our understanding of the relationship between a biomarker and dose of administered chemical. Several factors which need to be considered include: 1) the pharmacokinetics of the parent chemical, 2) the target tissue concentrations of the parent chemical or its bioactivated proximate toxicant, 3) the uptake kinetics of the parent chemical or metabolite into the target cell(s) and/or membrane interactions, and 4) the interaction of the chemical or metabolite with presumed receptor site(s). Because these theoretical factors each contain a saturable step due to definitive amounts of required enzyme, reuptake or receptor site(s), a nonlinear, saturable dose-response curve would be predicted. In order to exemplify this process, effects of the neurotoxicant, methlenedioxymethamphetamine (MDMA), were reviewed and analyzed. Our results and those of others indicate that: 1) peak concentrations of MDMA and metabolites are ochieved in rat brain by 30 min and are negligible by 24 hr, 2) a metabolite of MDMA is probably responsible for its neurotoxic effects, and 3) pretreatment with monoamine uptake blockers prevents MDMA neurotoxicity. When data generated from rats administerde MDMA were plotted as bilolgical effect (decreases in hippocampal serotonin concentrations) versus dose, a saturation curve best described the observed relationship. These results support the hypothesis that at least one saturable step is involved in MDMA neurotoxicity. We conclude that the mathematical relationship between biological effect and dose of MDMA, the first step of our quantitative neurotoxicity risk assessment procedure, should reflect this biological model information generated from the whole of the dose-response curve.