• The Journal of the Convergence on Culture Technology
• /
• v.9 no.4
• /
• pp.95-103
• /
• 2023

This paper proposes an artificial intelligence convergence education program for teaching the main concept and principle of Support Vector Machines(SVM) at elementary schools. The developed program, based on Jeju's natural environment theme, explains the decision boundary and margin of SVM by vertical and parallel from 4th grade mathematics curriculum. As a result of applying the developed program to 3rd and 5th graders, most students intuitively inferred the location of the decision boundary. The overall performance accuracy and rate of reasonable inference of 5th graders were higher. However, in the self-evaluation of understanding, the average value was higher in the 3rd grade, contrary to the actual understanding. This was due to the fact that junior learners had a greater tendency to feel satisfaction and achievement. On the other hand, senior learners presented more meaningful post-class questions based on their motivation for further exploration. We would like to find effective ways for artificial intelligence convergence education for elementary school students.

• Steel and Composite Structures
• /
• v.48 no.2
• /
• pp.207-233
• /
• 2023

Steel-concrete composite box girder bridges are widely used in the construction of highway and railway bridges both domestically and abroad due to their advantages of being light weight and having a large spanning ability and very large torsional rigidity. Composite box girder bridges exhibit the effects of shear lag, restrained torsion, distortion and interface bidirectional slip under various loads during operation. As one of the most commonly used calculation tools in bridge engineering analysis, one-dimensional models offer the advantages of high calculation efficiency and strong stability. Currently, research on the one-dimensional model of composite beams mainly focuses on simulating interface longitudinal slip and the shear lag effect. There are relatively few studies on the one-dimensional model which can consider the effects of restrained torsion, distortion and interface transverse slip. Additionally, there are few studies on vehicle-bridge integrated systems where a one-dimensional model is used as a tool that only considers the calculations of natural frequency, mode and moving load conditions to study the dynamic response of composite beams. Some scholars have established a dynamic analysis model of a coupled composite beam bridge-train system, but where the composite beam is only simulated using a Euler beam or Timoshenko beam. As a result, it is impossible to comprehensively consider multiple complex force effects, such as shear lag, restrained torsion, distortion and interface bidirectional slip of composite beams. In this paper, a 27 DOF vehicle rigid body model is used to simulate train operation. A two-node 26 DOF finite beam element with composed box beams considering the effects of shear lag, restrained torsion, distortion and interface bidirectional slip is proposed. The dynamic analysis model of the coupled composite box girder bridge-train system is constructed based on the wheel-rail contact relationship of vertical close-fitting and lateral linear creeping slip. Furthermore, the accuracy of the dynamic analysis model is verified via the measured dynamic response data of a practical composite box girder bridge. Finally, the dynamic analysis model is applied in order to study the influence of various mechanical effects on the dynamic performance of the vehicle-bridge system.

• The Journal of Bigdata
• /
• v.7 no.2
• /
• pp.31-41
• /
• 2022

Steel is one of the most fundamental components to mechanical industry. However, the quality of products are greatly impacted by the surface defects in the steel. Thus, researchers pay attention to the need for surface defects detector and the deep learning methods are the current trend of object detector. There are still limitations and rooms for improvements, for example, related works focus on developing the models but don't take into account real-time application with practical implication on industrial settings. In this paper, a real-time application of steel surface defects detection based on YOLOv4 is proposed. Firstly, as the aim of this work to deploying model on real-time application, we studied related works on this field, particularly focusing on one-stage detector and YOLO algorithm, which is one of the most famous algorithm for real-time object detectors. Secondly, using pre-trained Yolov4-Darknet platform models and transfer learning, we trained and test on the hot rolled steel defects open-source dataset NEU-DET. In our study, we applied our application with 4 types of typical defects of a steel surface, namely patches, pitted surface, inclusion and scratches. Thirdly, we evaluated YOLOv4 trained model real-time performance to deploying our system with accuracy of 87.1 % mAP@0.5 and over 60 fps with GPU processing.

• The Journal of Bigdata
• /
• v.8 no.1
• /
• pp.157-165
• /
• 2023

Seafood is a major source of protein in many countries and its consumption is increasing. In Korea, consumption of seafood is increasing, but self-sufficiency rate is decreasing, and the importance of safety management is increasing as the amount of imported seafood increases. There are hundreds of species of aquatic products imported into Korea from over 110 countries, and there is a limit to relying only on the experience of inspectors for safety management of imported aquatic products. Based on the data, a model that can predict the customs inspection results of imported aquatic products is developed, and a machine learning classification model that determines the non-conformity of aquatic products when an import declaration is submitted is created. As a result of customs inspection of imported marine products, the nonconformity rate is less than 1%, which is very low imbalanced data. Therefore, a sampling method that can complement these characteristics was comparatively studied, and a preprocessing method that can interpret the classification result was applied. Among various machine learning-based classification models, Random Forest and XGBoost showed good performance. The model that predicts both compliance and non-conformance well as a result of the clearance inspection is the basic random forest model to which ADASYN and one-hot encoding are applied, and has an accuracy of 99.88%, precision of 99.87%, recall of 99.89%, and AUC of 99.88%. XGBoost is the most stable model with all indicators exceeding 90% regardless of oversampling and encoding type.

• Korean Journal of Remote Sensing
• /
• v.39 no.2
• /
• pp.157-168
• /
• 2023

Image segmentation and supervised classification techniques are widely used to monitor the ground surface using high-resolution remote sensing images. In order to classify various objects, a process of defining a class corresponding to each object and selecting samples belonging to each class is required. Existing methods for extracting class samples should select a sufficient number of samples having similar intensity characteristics for each class. This process depends on the user's visual identification and takes a lot of time. Representative samples of the class extracted are likely to vary depending on the user, and as a result, the classification performance is greatly affected by the class sample extraction result. In this study, we propose an image classification technique that minimizes user intervention when extracting class samples by applying the histogram back-projection technique and has consistent intensity characteristics of samples belonging to classes. The proposed classification technique using histogram back-projection showed improved classification accuracy in both the experiment using hue subchannels of the hue saturation value transformed image from Compact Advanced Satellite 500-1 imagery and the experiment using the original image compared to the technique that did not use histogram back-projection.

• Journal of Service Research and Studies
• /
• v.12 no.1
• /
• pp.36-48
• /
• 2022

As the use of smartphones has rapidly increased due to the development of digital technology, the expansion of smartphones, and the COVID-19 incident, dependence on smartphones and the Internet is emerging as a serious social problem. As one of the solutions to the smartphone overdependence problem, the government and companies are releasing smartphone overdependence prevention applications. However, research on the effectiveness of smartphone overdependence prevention applications is insufficient. Therefore, this study selects 25 applications serviced in Korea as analysis targets and evaluates smartphone overdependence prevention applications in terms of function and service using the FGI survey method to identify problems and propose improvements. In the function evaluation, the functions of blocking illegal/harmful apps/websites, limiting smartphone usage time, and monitoring smartphone usage status are provided in most applications, so satisfaction scores are also highly evaluated. However, functions such as location check, smombie prevention, and body camphishing prevention served by some applications are evaluated low due to poor performance and poor accuracy. Classified by service provider, government-providing applications need to accurately perform functions and improve convenience of use. Mobile-Carrier-providing applications need to improve connectivity with other carriers and compatibility with other smart devices like smartphone, tablet, etc. Other private enterprise-providing applications need to open AS channels such as customer service centre and chatbot to improve service.

• Smart Media Journal
• /
• v.12 no.1
• /
• pp.9-16
• /
• 2023

Recently, a convolutional neural network (CNN) based system is being developed to overcome the limitations of human resources in the apple quality classification of farmhouse. However, since convolutional neural networks receive only images of the same size, preprocessing such as sampling may be required, and in the case of oversampling, information loss of the original image such as image quality degradation and blurring occurs. In this paper, in order to minimize the above problem, to generate a image patch based graph of an original image and propose a random walk-based positional encoding method to apply the graph transformer model. The above method continuously learns the position embedding information of patches which don't have a positional information based on the random walk algorithm, and finds the optimal graph structure by aggregating useful node information through the self-attention technique of graph transformer model. Therefore, it is robust and shows good performance even in a new graph structure of random node order and an arbitrary graph structure according to the location of an object in an image. As a result, when experimented with 5 apple quality datasets, the learning accuracy was higher than other GNN models by a minimum of 1.3% to a maximum of 4.7%, and the number of parameters was 3.59M, which was about 15% less than the 23.52M of the ResNet18 model. Therefore, it shows fast reasoning speed according to the reduction of the amount of computation and proves the effect.

• Analytical Science and Technology
• /
• v.19 no.5
• /
• pp.433-440
• /
• 2006

A HPLC/UV method was developed and validated for the rapid and simultaneous determination of urinary metabolites of organic solvents, mandelic acid, hippuric acid, phenylglyoxylic acid, ortho-, meta- and para-methylhippuric acid, using a monolithic column. The mobile phase was composed of tetrabutylammonium bromide as ion-pairing reagent with a flow rate of 2.4 mL/min. The total run time was less than 2.5 min for all six analytes. Good linearities were obtained for all the metabolites with correlation coefficients above 0.9993. Intra-day precision, accuracy and inter-day precision was 0.01~7.32%, 83.9~116.3% and 0.01~7.16%, respectively. The method was validated and confirmed by quantification of the quality assurance samples of Industrial Safety and Health Research Institute, Korea Occupational Safety and Health Agency.

• Analytical Science and Technology
• /
• v.20 no.4
• /
• pp.347-354
• /
• 2007

A simple and rapid high-performance liquid chromatography assay for the determination of residual novobiocin levels in bovine, porcine, chicken, flatfish and japanese eel muscle has been developed and validated. The separation condition for HPLC/UV was optimized with phenyl hexyl ($4.6{\times}150mm$, $5{\mu}m$) column with 10 mM monobasic sodium phosphate buffer (pH 2.5)/acetonitrile (50/50, v/v) as the mobile phase at a flow rate of 1.0 mL/min and detection wavelength was set at 254 nm. Residues were extracted from tissue by blending with methanol and lipid materials were removed with n-hexane. Then, the methanol extract was evaporated to dryness under a nitrogen stream, reconstituted in the mobile phase. Aliquot of the organic extract was decanted and filtered through $0.45{\mu}m$ syringe filter. The $20{\mu}L$ of the resulting solution was injected into the HPLC system. The calibration ranges were $0.5{\sim}5{\mu}g/g$ and calibration curves were linear with coefficients of correlation better than 0.95. The limits of quantification were $0.5{\mu}g/g$ for all muscles. The recoveries of bovine, porcine, chicken, flatfish and japaneseel muscles were 99.8%, 102.4%, 91.0%, 104.0% and 93.0%, respectively. The procedures were validated according to the CODEX guideline, determining specificity, linearity, accuracy, precision, quantitation limit and recovery.

• Analytical Science and Technology
• /
• v.22 no.1
• /
• pp.101-108
• /
• 2009

The bioequivalence of two pioglitazone tablets, Actos$^{(R)}$ tablet (Takeda Chemical Industries, reference drug) and Pioglitazone tablet (Boryung Company, test drug) was evaluated according to the guidelines of Korea Food and Drug Administration. Twenty-eight healthy male Korean volunteers received each medicine (pioglitazone dose of 30 mg) in a $2{\times}2$ crossover study with one week washout interval. After drug administration, blood samples were collected at specific time intervals from 0-36 hours. The plasma concentrations of pioglitazone were determined by high performance liquid chromatography-tandem mass spectrometry (LC-MS/MS). The total chromatographic run time was 5 min and calibration curves were linear over the concentration range of 5-2000 ng/mL for pioglitazone. The method was validated for selectivity, sensitivity, linearity, accuracy and precision. The pharmacokinetic parameters were determined from the plasma concentration-time profiles of both formulations. The primary calculated pharmacokinetic parameters were compared statistically to evaluate bioequivalence between the two preparations. The 90% confidence intervals of the $AUC_t$ ratio and the $C_{max}$ ratio for Pioglitazone tablet and Actos$^{(R)}$ tablet were log0.9422~log1.1040 and log0.9200~log1.1556, respectively. Based on the statistical considerations, we can conclude that the test drug, Pioglitazone tablet was bioequivalent to the reference drug, Actos$^{(R)}$ tablet.