• Journal of Welding and Joining
• /
• v.33 no.4
• /
• pp.24-29
• /
• 2015

The current waveform was analysed to monitor the weld quality in real time process. The acquired current waveform was discretely analysed for the top and bottom limits of peaks as well as the pulse frequency measurement. Fast Fourier Transform was implemented in the program to monitor the pulse frequency in real time. The developed algorithm or program was tested for the validation purpose. The cross-section of weld profile was compared to the current waveform profile to correlate the monitored signal and the actual parts. Pulse frequency was also used as auxiliary tool for the quality monitoring. Based on the results, it was possible to evaluate the quality of welding by measure the current waveform profile and frequency measurement.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.33 no.5
• /
• pp.529-535
• /
• 2009

This paper describes a measuring device of glass thickness using machine vision and image processing techniques on real-time. Today, the machine vision enable to inspect fast and exactly than human's eyes. The presented system has advantages of continuous measurement, flexibility and good accuracy. The system consists of a laser diode, a CCD camera with PC. The camera located on the opposite side of the incident beam measures the distance between two reflected laser beams from the glass top and bottom surface. We apply a binary algorithm to convert and analyze the image from camera to PC. Laser point coordination by border tracing algorithm is used to find the center of beam circle. The measured result was compared with micrometer and showed 0.002mm accuracy. Finally, the errors were discussed how to minimize the influence of glass wedge angle and angular error of moving stage.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.21 no.4
• /
• pp.225-244
• /
• 2017

The natural convective nanofluid flow and heat transfer inside a square enclosure with a center heater in the presence of magnetic field has been studied numerically. The vertical walls of the enclosure are cold and the top wall is adiabatic while the bottom wall is considered with constant heat source. The governing differential equations are solved by using a finite volume method based on SIMPLE algorithm. The parametric study is performed to analyze the effect of different lengths of center heater, Hartmann numbers and Rayleigh numbers. The heater effectiveness and temperature distribution are examined. The effect of all pertinent parameters on streamlines, isotherms, velocity profiles and average Nusselt numbers are presented. It is found that heat transfer increases with the increase of heater length, whereas it decreases with the increase of magnetic field effect. Furthermore, it is found that the value of Nusselt number depends strongly upon the Hartmann number for the increasing values of Rayleigh number.

• Journal of Broadcast Engineering
• /
• v.9 no.4 s.25
• /
• pp.446-454
• /
• 2004

With the growth of multimedia systems, compressing image data has become more important in the area of multimedia services. Since a compressed image bitstream can often be seriously distorted by various types of channel noise, an error concealment algorithm becomes a very important issue. In order to solve this problem, Hsia proposed the error concealment algorithm where he recovered lost block data using 1D boundary matching vectors. His algorithm, however, requires high computational complexity since each matching vector needs MAD (Mean Absolute Difference) values of all pixels, which is either a boundary line top or a boundary line bottom of a damaged block. We propose a hierarchical search-based fast error concealment scheme as well as its approximated version to reduce computational time. In the proposed scheme, a hierarchical search is applied to reduce the number of checking points for searching a vector. The error concealment schemes proposed in this paper can be about 3 times faster than Hsia's with keeping visual quality and PSNR.

• Smart Media Journal
• /
• v.2 no.2
• /
• pp.20-27
• /
• 2013

The concept of social media is top of the agenda for many business executives and decision makers, as well as consultants try to identify ways where companies can make profitable use of applications such as Netflix, Flixster. The social media is playing an increasingly important role as the information sources for customers making product choices etc. With the flourish of Web 2.0 technology, customer reviews are becoming more and more useful and important information resources for people to save their time and energy on purchasing products that they want. This paper proposes the Bayesian Probabilistic Classification algorithm to mine the social media review, and evaluates it by different splits and cross validation mechanism from the real data set. The explored study experimental results show the robustness and effectiveness of proposed approach for mining the social media review.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.7
• /
• pp.9-17
• /
• 2021

In this study, we propose a part of the CDSS(Clinical Decision Support System) study, a system that can classify chemotherapy, one of the treatment methods for colorectal cancer patients. In the treatment of colorectal cancer, the selection of chemotherapy according to the patient's condition is very important because it is directly related to the patient's survival period. Therefore, in this study, chemotherapy was classified using a machine learning algorithm by creating a baseline model, a pathological model, and a combined model using both characteristics of the patient using the individual and pathological characteristics of colorectal cancer patients. As a result of comparing the prediction accuracy with Top-n Accuracy, ROC curve, and AUC, it was found that the combined model showed the best prediction accuracy, and that the LGBM algorithm had the best performance. In this study, a chemotherapy classification model suitable for the patient's condition was constructed by classifying the model by patient characteristics using a machine learning algorithm. Based on the results of this study in future studies, it will be helpful for CDSS research by creating a better performing chemotherapy classification model.

• Proceedings of the KIEE Conference
• /
• 1998.07g
• /
• pp.2309-2311
• /
• 1998

The attitude control of a double inverted pendulum with a periodical disturbance at link top is dealt in this paper. The proposed system is consisted of the double inverted pendulum and a disturbance link. The lower link is hinged on the plate to free for rotation in the vertical plane. The upper link is connected to the lower link through a DC motor. The DC motor is used to control the posture of the pendulum by adjusting the position of the upper link. The periodical disturbance can be generated by the additional link attached at the end of link 2 through another DC motor, which is the modeling of a posture for a biped supporting with one leg. The motor for the joint simulates the knee joint(or hip joint) and the disturbance for the legs moving in air. The algorithm for controlling a proposed inverted pendulum is consisted of a state feedback control and a fuzzy logic controller. The fuzzy controller keeps the center of gravity of the biped within the specified range through the nonlinear feedback compensator. The state feedback control takes over the role to maintain a desired posture regardless the disturbance at the link top. In these case, the change of the angle and COG of an upper link is compensated with on-line. Simulations with a mathematical model are conducted to show the validity of the proposed controller.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.5
• /
• pp.495-507
• /
• 2019

Data sparsity is one of the main challenges for the recommender system. The recommender system contains massive data in which only a small part is the observed data and the others are missing data. Most studies assume that missing data is randomly missing from the dataset. Therefore, they only use observed data to train recommendation model, then recommend items to users. In actual case, however, missing data do not lost randomly. In our research, treat these missing data as negative examples of users' interest. Three sample methods are seamlessly integrated into SVD++ algorithm and then propose SVD++_W, SVD++_R and SVD++_KNN algorithm. Experimental results show that proposed sample methods effectively improve the precision in Top-N recommendation over the baseline algorithms. Among the three improved algorithms, SVD++_KNN has the best performance, which shows that the KNN sample method is a more effective way to extract the negative examples of the users' interest.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.1795-1811
• /
• 2019

In order to achieve rapid and accurate detection of vehicle objects in complex traffic conditions, we propose a novel vehicle detection method. Firstly, more contextual and small-object vehicle information can be obtained by our Joint Feature Network (JFN). Secondly, our Evolved Region Proposal Network (EPRN) generates initial anchor boxes by adding an improved version of the region proposal network in this network, and at the same time filters out a large number of false vehicle boxes by soft-Non Maximum Suppression (NMS). Then, our Mask Network (MaskN) generates an example that includes the vehicle occlusion, the generator and discriminator can learn from each other in order to further improve the vehicle object detection capability. Finally, these candidate vehicle detection boxes are optimized to obtain the final vehicle detection boxes by the Fine-Tuning Network(FTN). Through the evaluation experiment on the DETRAC benchmark dataset, we find that in terms of mAP, our method exceeds Faster-RCNN by 11.15%, YOLO by 11.88%, and EB by 1.64%. Besides, our algorithm also has achieved top2 comaring with MS-CNN, YOLO-v3, RefineNet, RetinaNet, Faster-rcnn, DSSD and YOLO-v2 of vehicle category in KITTI dataset.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.27 no.2
• /
• pp.256-264
• /
• 2024

In the last 10 years, AI has made rapid progress, and image classification, in particular, are showing excellent performance based on deep learning. Nevertheless, due to the nature of deep learning represented by a black box, it is difficult to actually use it in critical decision-making situations such as national defense, autonomous driving, medical care, and finance due to the lack of explainability of judgement results. In order to overcome these limitations, in this study, a model description algorithm capable of local interpretation was applied to the inception network-derived AI to analyze what grounds they made when classifying national defense data. Specifically, we conduct a comparative analysis of explainability based on confidence values by performing LIME analysis from the Inception v2_resnet model and verify the similarity between human interpretations and LIME explanations. Furthermore, by comparing the LIME explanation results through the Top1 output results for Inception v3, Inception v2_resnet, and Xception models, we confirm the feasibility of comparing the efficiency and availability of deep learning networks using XAI.