• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.11
• /
• pp.2037-2042
• /
• 2017

The manufacturing facility is generally operated by a pre-set program under the existing factory automation system. On the other hand, the manufacturing facility must decide how to operate autonomously in Industry 4.0. Determining the operation mode of the production facility itself means, for example, that it detects the abnormality such as the deterioration of the facility at the shop-floor, prediction of the occurrence of the problem, detection of the defect of the product, In this paper, we propose a manufacturing process modeling using a queue for detection of manufacturing process abnormalities at the shop-floor, and detect abnormalities in the modeling using SVM, one of the machine learning techniques. The queue was used for M / D / 1 and the conveyor belt manufacturing system was modeled based on ${\mu}$, ${\lambda}$, and ${\rho}$. SVM was used to detect anomalous signs through changes in ${\rho}$.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.3
• /
• pp.171-179
• /
• 2020

Recently, studies have shown that convolution neural networks are achieving the best performance in image classification, object detection, and image generation. Vision based defect inspection which is more economical than other defect inspection, is a very important for a factory automation. Although supervised anomaly detection algorithm has far exceeded the performance of traditional machine learning based method, it is inefficient for real industrial field due to its tedious annotation work, In this paper, we propose ADGAN, a unsupervised anomaly detection architecture using the variational autoencoder and the generative adversarial network which give great results in image generation task, and demonstrate whether the proposed network architecture identifies anomalous images well on MNIST benchmark dataset as well as our own welding defect dataset.

• Journal of IKEEE
• /
• v.24 no.2
• /
• pp.475-485
• /
• 2020

Recently, a crowdsensing system that provides a new sensing service with real-time sensing data provided from a user's device including a sensor without installing a separate sensor has attracted attention. In the crowdsensing system, meaningless data may be provided due to a user's operation error or communication problem, or false data may be provided to obtain compensation. Therefore, the detection and removal of the abnormal data determines the quality of the crowdsensing service. The proposed methods in the past to detect these anomalies are not efficient for the fast-changing environment of crowdsensing. This paper proposes an anomaly data detection method by extracting the characteristics of continuously and rapidly changing sensing data environment by using machine learning technology and modeling it with an appropriate algorithm. We show the performance and feasibility of the proposed system using deep learning binary classification model of supervised learning and autoencoder model of unsupervised learning.

• Korean Journal of Veterinary Research
• /
• v.55 no.2
• /
• pp.97-103
• /
• 2015

In this study, the Korean veterinary medical devices management system was evaluated relative to systems in the USA, EU, and Japan. Veterinary medical devices are regulated in Korea based on the Medical Appliance Act of 1997. This was initially supervised by the Ministry of Agriculture, Food and Rural Affairs and Korea Animal Health Products Association, and subsequently by the Animal and Plant Quarantine Agency (QIA) in 2000. These devices were classified approximately 1,400 categories as instruments, supplies, artificial insemination apparatus, and other categories. Each of these devices was assigned to four regulatory grades by the QIA in 2007. The ranking system for veterinary medical devices was implemented in 2014 with 820 products from 162 companies registered by that year. However, in vitro diagnostic devices (IVDDs) for animals were managed as medical devices and biological medicine. In vitro diagnostic reagents for treating infection diseases are not subjected to either a classification or grading system. Veterinary medical devices are currently exempt from good manufacturing practices (GMP) and device tracking requirements. Due to gradual growth of the domestic veterinary medical devices market since 2008, regulation of these devices should be improved with re-examination of IVDDs and GMP certification for the effective operating system.

• Journal of the Korean Society of Food Science and Nutrition
• /
• v.43 no.3
• /
• pp.419-424
• /
• 2014

Metabolite profiling of blueberry (cultivar "Spartan") was performed by extraction using different solvents, methanol and ethyl acetate, through metabolomic analysis using LC-MS/MS. Unsupervised classification method (PCA) and supervised prediction model (OPLS-DA) provided good categorization of metabolites according to the extraction solvents. Metabolites of the anthocyanin family, including delphinidin hexoside, delphinidin, 5-O-feruloylquinic acid, malvidin hexoside, malvidin-3-arabinoside, petunidin-3-arabinoside, and petunidin hexoside, were mainly detected in methanol fractions, whereas those of the flavonoid family, including chlorogenic acid, chlorogenic acid dimer, 6,8-di-C-arabinopyranosyl-luteolin, and luteolin were successfully prepared in the ethyl acetate fraction. Thus, metabolomic analysis of blueberry extracts allows for the simple profiling of whole and distinctive metabolites for future applications.

• Progress in Medical Physics
• /
• v.30 no.2
• /
• pp.49-58
• /
• 2019

Purpose: Automated analytical systems have begun to emerge as a database system that enables the scanning of medical images to be performed on computers and the construction of big data. Deep-learning artificial intelligence (AI) architectures have been developed and applied to medical images, making high-precision diagnosis possible. Materials and Methods: For diagnosis, the medical images need to be labeled and standardized. After pre-processing the data and entering them into the deep-learning architecture, the final diagnosis results can be obtained quickly and accurately. To solve the problem of overfitting because of an insufficient amount of labeled data, data augmentation is performed through rotation, using left and right flips to artificially increase the amount of data. Because various deep-learning architectures have been developed and publicized over the past few years, the results of the diagnosis can be obtained by entering a medical image. Results: Classification and regression are performed by a supervised machine-learning method and clustering and generation are performed by an unsupervised machine-learning method. When the convolutional neural network (CNN) method is applied to the deep-learning layer, feature extraction can be used to classify diseases very efficiently and thus to diagnose various diseases. Conclusions: AI, using a deep-learning architecture, has expertise in medical image analysis of the nerves, retina, lungs, digital pathology, breast, heart, abdomen, and musculo-skeletal system.

• Clinics in Shoulder and Elbow
• /
• v.1 no.2
• /
• pp.198-204
• /
• 1998

Unstable fractures of the proximal humerus continue to be difficult problems for orthopaedic surgeons. The optimum treatment of these fractures has remained a matter of controversy. We analyzed the clinical results of open reduction and plate fixation underwent for patients of unstable fractures of proximal humerus after minimum 12 months follow up. The purpose of this study is to evaluate the efficacy of open reduction and rigid plate fixation. Twenty-two patients were managed with open reduction and plate fixation. Mean follow up duration was 20.6 months(range, 12 to 28 mon.). Because the age of patient as a maker of degree of osteoporosis was considered the key factor in the success of anatomic reconstruction, we divided into two groups according to age. Group A was comprised of 12 cases with younger than 50 yrs of age. Ten cases of older than 50 yrs of age were Group B. According to Neer's classification, five cases(22%) were two part fracture, 12 cases(64%) were three part fracture, and three cases(14%) were four part fracture. We used the Neer rating system for evaluating the results. In Group A, overall scores were 79.1. In Group B, overall scores were 76.8. Overall scores in two part fracture were 85, overall scores in three part fracture 78.4 and overall scores in three part fracture 68.3. We achieved excellent or good results in nine cases(75%) of Group A and seven cases(70%) of Group B. Also, we obtained excellent or good results in all cases of two part fracture, ten cases(71%) of three fracture and one case(33%) of four part fracture. The complications were three metal loosening, one avascular necrosis of humeral head, one severe stiff shoulder, one superficial wound infection and one ectopic ossification. The results were excellent or good in 16 cases(73%) out of 22 cases. In conclusion, rigid fixation and supervised early exercise would be a good option for unstable fracture of the proximal humerus.

• Proceedings of the KSRS Conference
• /
• 2007.10a
• /
• pp.426-429
• /
• 2007

One of the widely used applications of remote sensing studies is environmental change detection and biodiversity conservation. The study area Hustai Mountain is situated in the transition zone between the Siberian taiga forest and Central Mongolian arid steppe. Hustai National Park carries out one of several reintroduction programs of takhi (wild horse or Equus ferus przewalskii) from various zoos in the world and it represents one of a few textbook examples of successful reintroduction of an animal extinct in the wild. In this paper we describe the results of an analysis on the change of remaining forest area over the 7-year period since Hustai Mountain was designated as a protected area for reintroduction to wild horses. Today the forested area covers approximately 5% of the Hustai National Park, mostly the north-facing slopes above 1400 m altitude. Birch (Betula platyphylla) and aspen (Populus tremula) trees are predominant in the forest. We used Landsat ETM+ images from two different years and multi temporal MODIS NDVI data. Land types were determined by supervised classification methods (Maximum Likelihood algorithm) verified with ground-truthing data and the Land Change Modeler (LCM) which was developed by Clark Labs. Forested area was classified into three different land types, namely the forest land, mountain meadow and mountain steppe. The study results illustrate that the remaining birch forest has rapidly changed to fragmented forest land and to open areas. Underlying causes for such a rapid change during the 15-year period may be manifold. However, the responsible factors appear to be the drying off and outbreak of forest pest species (such as gypsy moth or Lymantria dispar) in the area.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.46 no.7
• /
• pp.3-12
• /
• 2004

The need to predict the rate of soil erosion, both under existing conditions and those expected to occur following soil conservation practice, has been led to the development of various models. In this study Morgan model especially developed for field-sized areas on hill slopes was applied to assess the rate of soil erosion using RS/GIS environment in the Dukchun river basin, one of two tributaries flowing into Jinyang lake. In order to run the model, land cover mapping was made by the supervised classification method with Landsat TM satellite image data, the digital soil map was generated from scanning and screen digitizing from the hard copy of soil maps, digital elevation map (DEM) in order to generate the slope map was made by the digital map (DM) produced by National Geographic Information Institute (NGII). Almost all model parameters were generated to the multiple raster data layers, and the map calculation was made by the raster based GIS software, IL WIS which was developed by ITC, the Netherlands. Model results show that the annual soil loss rates are 5.2, 18.4, 30.3, 58.2 and 60.2 ton/ha/year in forest, paddy fields, built-up area, bare soil, and upland fields respectively. The estimated rates seemed to be high under the normal climatic conditions because of exaggerated land slopes due to DEM generation using 100 m contour interval. However, the results were worthwhile to estimate soil loss in hilly areas and the more precise result could be expected when the more accurate slope data is available.

• Journal of Digital Convergence
• /
• v.11 no.11
• /
• pp.409-414
• /
• 2013

Naive Bayesian classifiers are a powerful and well-known type of classifiers that can be easily induced from a dataset of sample cases. However, the strong conditional independence assumptions can sometimes lead to weak classification performance. Normally, naive Bayesian classifiers use Gaussian distributions to handle continuous attributes and to represent the likelihood of the features conditioned on the classes. The probability density of attributes, however, is not always well fitted by a Gaussian distribution. Another eminent type of classifier is the neuro-fuzzy classifier, which can learn fuzzy rules and fuzzy sets using supervised learning. Since there are specific structural similarities between a neuro-fuzzy classifier and a naive Bayesian classifier, the purpose of this study is to apply learning distribution graphs constructed by a neuro-fuzzy network to naive Bayesian classifiers. We compare the Gaussian distribution graphs with the fuzzy distribution graphs for the naive Bayesian classifier. We applied these two types of distribution graphs to classify leukemia and colon DNA microarray data sets. The results demonstrate that a naive Bayesian classifier with fuzzy distribution graphs is more reliable than that with Gaussian distribution graphs.