• Journal of Internet Computing and Services
• /
• v.24 no.6
• /
• pp.91-98
• /
• 2023

Recently, according to development of artificial intelligence, a wide range of industry being automatic and optimized. Also we can find out some research of using supervised learning for deteceting defect of railway in domestic rail industry. However, there are structures other than rails on the track, and the fastener is a device that binds the rail to other structures, and periodic inspections are required to prevent safety accidents. In this paper, we present a method of reducing cost for labeling using semi-supervised and transfer model trained on rail fastener data. We use Resnet50 as the backbone network pretrained on ImageNet. At first we randomly take training data from unlabeled data and then labeled that data to train model. After predict unlabeled data by trained model, we adopted a method of adding the data with the highest probability for each class to the training data by a predetermined size. Futhermore, we also conducted some experiments to investigate the influence of the number of initially labeled data. As a result of the experiment, model reaches 92% accuracy which has a performance difference of around 5% compared to supervised learning. This is expected to improve the performance of the classifier by using relatively few labels without additional labeling processes through the proposed method.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.4
• /
• pp.50-58
• /
• 2017

Recently, some methods of converging heterogeneous fire sensor data have been proposed for effective fire detection, but the rule-based methods have low adaptability and accuracy, and the fuzzy inference methods suffer from detection speed and accuracy by lack of consideration for images. In addition, a few image-based deep learning methods were researched, but it was too difficult to rapidly recognize the fire event in absence of cameras or out of scope of a camera in practical situations. In this paper, we propose a novel fire detection system combining a deep learning algorithm based on CNN and fuzzy inference engine based on heterogeneous fire sensor data including temperature, humidity, gas, and smoke density. we show it is possible for the proposed system to rapidly detect fire by utilizing images and to decide fire in a reliable way by utilizing multi-sensor data. Also, we apply distributed computing architecture to fire detection algorithm in order to avoid concentration of computing power on a server and to enhance scalability as a result. Finally, we prove the performance of the system through two experiments by means of NIST's fire dynamics simulator in both cases of an explosively spreading fire and a gradually growing fire.

• Progress in Superconductivity
• /
• v.9 no.1
• /
• pp.23-28
• /
• 2007

In order to have a superconductive connection between the wire-wound pickup coil and input coil, typically Nb terminal blocks with screw holes are used. Since this connection structure occupies large volume, large stray pickup area can be generated which can pickup external noise fields. Thus, SQUID and connection block are shielded inside a superconducting tube, and this SQUID module is located at some distance from the distal coil of the gradiometer to minimize the distortion or imbalance of uniform background field due to the superconducting module. To operate this conventional SQUID module, we need a higher liquid He level, resulting in shorter refill interval. To make the fabrication of gradiometers simpler and refill interval longer, we developed a novel method of connecting the pickup coil into the input coil. Gradiometer coil wound of 0.125-mm diameter NbTi wires were glued close to the input coil pads of SQUID. The superconductive connection was made using an ultrasonic bonding of annealed 0.025-mm diameter Nb wires, bonded directly on the surface of NbTi wires where insulation layer was stripped out. The reliability of the superconductive bonding was good enough to sustain several thermal cycling. The stray pickup area due to this connection structure is about $0.1\;mm^2$, much smaller than the typical stray pickup area using the conventional screw block method. By using this compact connection structure, the position of the SQUID sensor is only about 20-30 mm from the distal coil of the gradiometer. Based on this compact module, we fabricated a magnetocardiography system having 61 first-order axial gradiometers, and measured MCG signals. The gradiometers have a coil diameter of 20 mm, and the baseline is 70 mm. The 61 axial gradiometer bobbins were distributed in a hexagonal lattice structure with a sensor interval of 26 mm, measuring $dB_z/dz$ component of magnetocardiography signals.

• Conservation Science in Museum
• /
• v.24
• /
• pp.1-16
• /
• 2020

This paper presents the process of re-conservation and the results of research on the production method of the Iron Sword with Ring Pommel with Three-Pointed Leaf Decoration, one of the excavation findings from Tomb No. 55 in the Dalseong Ancient Tomb Complex in Daegu. This iron sword is a double weapon with two large swords housed within a single sheath. Four smaller swords are attached to the surface of the sheath, two on the upper portion and the other two below. It is the only such two-in-one weapon excavated intact thus far from an ancient Korean tomb. The records show that it underwent conservation treatment twice in the past. In this study, it was subjected to conservation treatment again to replace the stiffener in some cracking areas, and its material, composition, and production method were analyzed by CT, XRF analysis, and stereoscopic microscopy. The sword is mainly made of copper, but the golden component contains both gold and mercury, which suggests that the copper was plated in gold using mercury amalgamation. The examination of the production methods indicates that it was intended more to demonstrate the authority of its owner rather than to serve any practical use. The two upper small swords on the sheath were made in the same manner as the main swords, and the two small lower swords were cut from a single metal sheet. The sheath was made by cutting two metal sheets. Supports were used to attach the two small swords to the upper portion of the sheath, while the lower portion of the sheath was slit to allow the other two small swords to be inserted into it. The ring pommels of the main swords have a three-pointed leaf decoration. As for the other designs, the handle of the main sword features a series of semicircles, the decorative bands on the sheath have waves in dots, and the fish tail of the sheath shows diagonal lines of dots.