• IEMEK Journal of Embedded Systems and Applications
• /
• v.19 no.3
• /
• pp.151-158
• /
• 2024

Although on-device artificial intelligence (AI) has gained attention to diagnosing machine faults in real time, most previous studies did not consider the model retraining and redeployment processes that must be performed in real-world industrial environments. Our study addresses this challenge by proposing an on-device AI-based real-time machine fault diagnosis system that utilizes continual learning. Our proposed system includes a lightweight convolutional neural network (CNN) model, a continual learning algorithm, and a real-time monitoring service. First, we developed a lightweight 1D CNN model to reduce the cost of model deployment and enable real-time inference on the target edge device with limited computing resources. We then compared the performance of five continual learning algorithms with three public bearing fault datasets and selected the most effective algorithm for our system. Finally, we implemented a real-time monitoring service using an open-source data visualization framework. In the performance comparison results between continual learning algorithms, we found that the replay-based algorithms outperformed the regularization-based algorithms, and the experience replay (ER) algorithm had the best diagnostic accuracy. We further tuned the number and length of data samples used for a memory buffer of the ER algorithm to maximize its performance. We confirmed that the performance of the ER algorithm becomes higher when a longer data length is used. Consequently, the proposed system showed an accuracy of 98.7%, while only 16.5% of the previous data was stored in memory buffer. Our lightweight CNN model was also able to diagnose a fault type of one data sample within 3.76 ms on the Raspberry Pi 4B device.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.1
• /
• pp.453-458
• /
• 2024

Due to global issues such as climate crisis and rising energy costs, there is an increasing focus on energy conservation and management. In the case of South Korea, approximately 53.5% of the total energy consumption comes from industrial complexes. In order to address this, we aimed to improve issues through the 'Shared Network Utility Plant' among companies using similar energy utilities to find energy-saving points. For effective energy conservation, various techniques are utilized, and stable data supply is crucial for the reliable operation of factories. Many anomaly detection and alert systems for checking the stability of data supply were dependent on Energy Management Systems (EMS), which had limitations. The construction of an EMS involves large-scale systems, making it difficult to implement in small factories with spatial and energy constraints. In this paper, we aim to overcome these challenges by constructing a data collection system and anomaly detection alert system on embedded devices that consume minimal space and power. We explore the possibilities of utilizing anomaly detection alert systems in typical institutions for data collection and study the construction process.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.1
• /
• pp.471-476
• /
• 2024

With the spread of various smart devices and computing devices, big data generation is occurring widely. Machine learning is an algorithm that performs reasoning by learning data patterns. Among the various machine learning algorithms, the algorithm that attracts attention is deep learning based on neural networks. Deep learning is achieving rapid performance improvement with the release of various applications. Recently, among deep learning algorithms, attempts to analyze data using graph structures are increasing. In this study, we present a graph generation method for transferring to a deep learning network. This paper proposes a method of generalizing node properties and edge weights in the graph generation process and converting them into a structure for deep learning input by presenting a matricization We present a method of applying a linear transformation matrix that can preserve attribute and weight information in the graph generation process. Finally, we present a deep learning input structure of a general graph and present an approach for performance analysis.

• Journal of the Korea Institute of Building Construction
• /
• v.9 no.5
• /
• pp.79-86
• /
• 2009

In a steel structure, spray-applied fire resistive materials are crucial in preventing structural strength from being weakened in the event of a fire. The quality control of such materials, however, is difficult for manual workers, who can frequently be in short supply. These skilled workers are also very likely to be exposed to environmental hazards. Problems with construction work such as this, which are specifically the difficulty of achieving quality control and the dangerous nature of the work itself, can be solved to some degree by the introduction of automated equipment. It is, however, very difficult to automate the work process, from operation to the selection of a location for the equipment, as the environment of a construction site has not yet been structured to accommodate automation. This is a fundamental study on the possibility of the automation of spray-applied fire resistive coating work. In this study, the linkability of the cutting-edge RTLS to an automation system is reviewed, and a scenario for the automation of spray-applied fire resistive coating work and system composition is presented. The system suggested in this study is still in a conceptual stage, and as such, there are many restrictions still to be resolved. Despite this fact, automation is expected to have good effectiveness in terms of preventing fire from spreading by maintaining a certain level of strength at a high temperature when a fire occurs, as it maintains the thickness of the fire-resistive coating at a specified level, and secures the integrity of the coating with the steel structure, thereby enhancing the fire-resistive performance. It also expected that if future research is conducted in this area in relation to a cutting-edge monitoring TRS, such as the ubiquitous sensor network (USN) and/or building information model (BIM), it will contribute to raising the level of construction automation in Korea, reducing costs through the systematic and efficient management of construction resources, shortening construction periods, and implementing more precise construction

• 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%.

• Journal of Distribution Science
• /
• v.12 no.2
• /
• pp.95-103
• /
• 2014

Purpose - Korea has a better geographical location than other nations in the Northeast Asian region. This means that Korea has an opportunity to become the center of international physical distribution in Northeast Asia. Korea should take advantage of this opportunity by exploring appropriate strategies to achieve this goal, assuming government willingness, with a view to capitalizing on the geographical advantage of the Korean peninsula and constructing a comprehensive physical distribution network system. If we prepare for this scenario, Korea could become the center of international physical distribution in Northeast Asia. Research design, data, and methodology - This study has the purpose of determining how shipping companies form partnerships with third-party logistics providers, and the relevant implications. The survey methods used were personal interview and a questionnaire distributed through e-mail, fax, mail, and telephone. A total of 600 questionnaires were distributed, out of which 285 were returned. Of the collected questionnaires, 10 were excluded because of insufficient content, leaving 275 to be used in the study as available valid samples. The data that was collected from these samples was analyzed using the data coating process and by employing a statistical package program. Results - Flexible policies, administration, and systems will be needed to create better business practices. In this dissertation, first and foremost, the results reveal that in order to become the center of Northeast Asian logistics, Korea must transition into a new paradigm based on the current economic and social systems that have stemmed from bureaucracy, inflexibility, chauvinism, and egalitarianism. Flexible policies, administration, and systems will be required to create better business practices. Domestic logistics corporations need to occupy a strategic logistics hub, create a logistics network, and activate value-added logistics business strategies by ensuring significant manpower and by building a logistics information system to strengthen their competitive edge, creating an improved system. Conclusions - In this dissertation, first and foremost, we point out that in order to become a center of North East Logistics, Korea should change to a new paradigm from the old one based on current economic and social systems that have stemmed from bureaucracy, inflexibility, chauvinism, and egalitarianism. More reasonable business laws, systems, and policies based on market-driven flexibility and transparency should be created. Moreover, social norms and rules should be reasonably established, to accomplish political and social security. Korea has to cultivate a culture of tolerance for foreign companies. This involves a change of paradigm for the development of the capital city and satellite cities. It will take a powerful task force or organization to plan and execute the vision that aims to meet these needs, accomplish the necessary goals, use the appropriate system effectively, and make Korea a key country in the field of Northeast Asian logistics.

• Journal of Internet Computing and Services
• /
• v.19 no.5
• /
• pp.43-54
• /
• 2018

Images taken in haze weather are characteristic of low contrast and poor visibility. The process of reconstructing clear-weather image from a hazy image is called dehazing. The main challenge of image dehazing is to estimate the transmission map or depth map for an input hazy image. In this paper, we propose a single image dehazing method by utilizing the Generative Adversarial Network(GAN) for accurate depth map estimation. The proposed GAN model is trained to learn a nonlinear mapping between the input hazy image and corresponding depth map. With the trained model, first the depth map of the input hazy image is estimated and used to compute the transmission map. Then a guided filter is utilized to preserve the important edge information of the hazy image, thus obtaining a refined transmission map. Finally, the haze-free image is recovered via atmospheric scattering model. Although the proposed GAN model is trained on synthetic indoor images, it can be applied to real hazy images. The experimental results demonstrate that the proposed method achieves superior dehazing results against the state-of-the-art algorithms on both the real hazy images and the synthetic hazy images, in terms of quantitative performance and visual performance.

• The Transactions of the Korea Information Processing Society
• /
• v.5 no.5
• /
• pp.1274-1284
• /
• 1998

Multicasting plays an important role in providing multimedia services like audio and video conferencing and has been actively studied for many years. The study of multicasting has been mainly focused on finding a low-cost tree on the assumption that every edge of the tree has enough link capacity to be allocated to the amound of bandwidth required by a service. In some cases, however, a multicasting tree whose some edges do not have enough bandwidth to be allocated may exist. A typical example of this case is to multicast among the group users located in the different sub-networks in the interconnected networks composed of the high bandwidth networks such as B-ISDN and the low bandwidth networks such as PSTN and PSDN. This paper proposes a novel multicasting protocol for providing multimedia services in a wide area network interconnected by multiple heterogeneous networks. First, the paper discusses the inherent limitation of present multicasting protocols for the case. Second, se introduce a new concept, multiple QoS's for a service, which extends the implicit assumption that only one QoS is assigned to a service in order to overcome this limitation. Third, our multicasting protocol is designed based on the current Internet protocol stacks. Last, we perform simulations for possible scenarios to provide multimedia services in the network and analyze their pros and cons.

• Journal of the Institute of Convergence Signal Processing
• /
• v.10 no.1
• /
• pp.7-12
• /
• 2009

In this paper, we propose a directional feature extraction method for off-line handwritten numerals by using the morphological operations. Direction features are obtained from four directional line images, each of which contains horizontal, vertical, right-diagonal and left-diagonal lines in entire numeral lines. Conventional method for extracting directional features uses Kirsch masks which generate edge-shaped double line images for each direction, whereas our method uses directional erosion operations and generate single line images for each direction. To apply these directional erosion operations to the numeral image, preprocessing steps such as thinning and dilation are required, but resultant directional lines are more similar to numeral lines themselves. Our four [$4{\times}4$] directional features of a numeral are obtained from four directional line images through a zoning method. For obtaining the higher recognition rates of the handwrittern numerals, we use the multiple feature which is comprised of our proposed feature and the conventional features of a kirsch directional feature and a concavity feature. For recognition test with given features, we use a multi-layer perceptron neural network classifier which is trained with the back propagation algorithm. Through the experiments with the CENPARMI numeral database of Concordia University, we have achieved a recognition rate of 98.35%.

• Journal of IKEEE
• /
• v.21 no.4
• /
• pp.412-415
• /
• 2017

In this paper, we propose a multi license plate recognition system using high resolution $360^{\circ}$ omnidirectional IP camera. The proposed system consists of a planar division part of $360^{\circ}$ circular image and a multi license plate recognition part. The planar division part of the $360^{\circ}$ circular image are divided into a planar image with enhanced image quality through processes such as circular image acquisition, circular image segmentation, conversion to plane image, pixel correction using color interpolation, color correction and edge correction in a high resolution $360^{\circ}$ omnidirectional IP Camera. Multi license plate recognition part is through the multi-plate extraction candidate region, a multi-plate candidate area normalized and restore, multiple license plate number, character recognition using a neural network in the process of recognizing a multi-planar imaging plates. In order to evaluate the multi license plate recognition system using the proposed high resolution $360^{\circ}$ omnidirectional IP camera, we experimented with a specialist in the operation of intelligent parking control system, and 97.8% of high plate recognition rate was confirmed.