• International Journal of Internet, Broadcasting and Communication
• /
• v.9 no.1
• /
• pp.9-17
• /
• 2017

This paper describes a process of developing commercial real time image recognition system with company. In this paper we will make a system that is combining an empirical kernel map method and conjugate least squares support vector machine in order to represent images in a low-dimensional subspace for real time image recognition. In the traditional approach calculating these eigenspace models, known as traditional PCA method, model must capture all the images needed to build the internal representation. Updating of the existing eigenspace is only possible when all the images must be kept in order to update the eigenspace, requiring a lot of storage capability. Proposed method allows discarding the acquired images immediately after the update. By experimental results we can show that empirical kernel map has similar accuracy compare to traditional batch way eigenspace method and more efficient in memory requirement than traditional one. This experimental result shows that proposed model is suitable for commercial real time image recognition system.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2006.05a
• /
• pp.635-636
• /
• 2006

A knowledge-based system with CAD is widely used to implement a designer's know-how and a routine calculation within a system. An engineer needs to change several design parameters in an early design stage. However it is not easy to change the design parameters when it needs to update an existing model and data. In this paper, a module has been proposed to make it easy to change the design parameter so that the engineer can easily update and change the design model. Also a flexible interface of the module can easily add or extract the design knowledge.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.20 no.3
• /
• pp.313-321
• /
• 2002

Supplying high-resolution satellite image, we take much interest in the update and the revision of digital map and thematic map based on the satellite image. This study presented the possibility of the update and the revision to the existing digital map on a scale of l/5,000 and 1/25,000 to take advantage of the IKONOS satellite image. We performed geometric correction to make use of the ground control points of the existing digital map in IKONOS mono-image and created ortho-image by extracting digital elevation model from three dimensional contour data and altitude on the existing digital map. We revised changed features in the method of screen digitizing by overlapping orthorectified satellite image and existing digital map and flawed features of the unchanged area on the satellite images for positional accuracy analysis. As a result, rectification error is calculated at $\pm$3.35m by RMSE. There is a good possibility of update of digital map under the scale of 1/10,000. It is possible to the update of the large scale digital map over the scale of l/5,000, as if we used the method of stereo image and ground control point surveying.

• Proceedings of the Korean Radioactive Waste Society Conference
• /
• 2018.11a
• /
• pp.217-218
• /
• 2018

• Korean Management Science Review
• /
• v.26 no.3
• /
• pp.23-35
• /
• 2009

This study deals with the location management in WCDMA (Wideband Code Division Multiple Access) networks. The new location management method is the one of the many techniques used to provide the packet switched (PS) services effectively in WCDMA networks. WCDMA introduces RA (routing area) for the PS services in addition to LA (location area) for the circuit switched (CS) services. WCDMA system also adopts the cooperative paging and the combined area update which are to decrease signaling traffic using the PS network resources for the CS services. Considering the characteristics of the WCDMA networks, this study formulate into the mathematical programming problem for the location management. Rather than the existing researches which consider the single MSC/SGSN networks, we deal with the multi MSC/SGSN networks, where both MSC area and SGSN area should be determined as well. Fairness between traffic loads of MSC/SGSN and the system cost incurred to MSC/SGSN are also incorporated into our model, which make the model more realistic. We propose greedy algorithms for the problems, which consists of the merger of two neighboring areas and the movement of an area. Extensive experiment has been done based on the realistic problem examples. Followed by the analysis of parameter effect on the final solutions and the quality of final solutions obtained by our greedy algorithms. Our proposed model and analysis result can be used to determine WCDMA system areas and to predict the system performance measures by the determination.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.7 no.3
• /
• pp.182-187
• /
• 2007

A fuzzy modeling method is proposed to build the dynamic model of a walking-beam reheating furnace from the recorded data. In the proposed method, the number of membership function on each variable is increased individually and the modeling accuracy is evaluated iteratively. When the modeling accuracy is satisfied, the membership functions on each variable are fixed and the structure of fuzzy model is determined. Because the training data is limited, in this process, as the number of membership function increase, it is highly possible that some rules are missing, i.e., no data in the training set corresponds to the consequent part of a missing rule. To complete the rulebase, the output of the model constructed at the previous step is used to generate the consequent part of the missing rules. Finally, in the real time application, a rolling update scheme to rulebase is introduced to compensate the change of system dynamics and fine tune the rulebase. The proposed method is verified by the application to the modeling of a reheating furnace.

• Journal of the Institute of Convergence Signal Processing
• /
• v.8 no.4
• /
• pp.297-306
• /
• 2007

Model reference control is control method such that overall response of a plant plus controller approaches that of a given reference model. The reference model provides desired trajectory the plant should follow. There are many kinds of control methods in MRC. However, this paper focuses on Model Reference Sliding Mode Control. The plant of these controls is an uncertain and linear system varying in time, of second order, and with SISO. In this paper, a design scheme of reference model is proposed for MRSMC. The scheme determines reference model based on the information on bounded control inputs, reference inputs and system parameters. It is used to choose a Fixed Reference Model in the process of controller design, to update Variable Reference Model when stepwise reference inputs change and to update Instant Reference Model at every sampling time. The simulation results show that the proposed method yields better control performance than the conventional MRC subject to the stepwise reference input when applied to the position control system for motor system.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.23 no.4
• /
• pp.1-15
• /
• 2020

As cities expand and underground utilities construction projects increase, there is an urgent need for a technology capable of analyzing the underground utilities network in 3D. Since 2015, 3D Underground Geospatial Map project, that has been integrating 15 types of underground information such as underground utilities, underground structures, and ground information, is in progress in S. Korea. However, the construction of 3D underground facilities is currently based on manual work and the logic for building a 3D model is very complicated. And it takes a lot of time and cost to process millions of large amounts of data per local governments. By presenting a framework on the processing and partial updating of the 3D underground utilities model, this paper aims to establish a plan to quickly build a 3D underground utility model at a minimum cost. The underground utilities processing and partial update automation technologies developed in this study are expected to be immediately applied to the 3D Underground Geospatial Map project.

• Journal of Information Processing Systems
• /
• v.8 no.1
• /
• pp.55-66
• /
• 2012

This paper presents a dual modeling method that simulates the graphic and haptic behavior of a volumetric deformable object and conveys the behavior to a human operator. Although conventional modeling methods (a mass-spring model and a finite element method) are suitable for the real-time computation of an object's deformation, it is not easy to compute the haptic behavior of a volumetric deformable object with the conventional modeling method in real-time (within a 1kHz) due to a computational burden. Previously, we proposed a fast volume haptic rendering method based on the S-chain model that can compute the deformation of a volumetric non-rigid object and its haptic feedback in real-time. When the S-chain model represents the object, the haptic feeling is realistic, whereas the graphical results of the deformed shape look linear. In order to improve the graphic and haptic behavior at the same time, we propose a dual modeling framework in which a volumetric haptic model and a surface graphical model coexist. In order to inspect the graphic and haptic behavior of objects represented by the proposed dual model, experiments are conducted with volumetric objects consisting of about 20,000 nodes at a haptic update rate of 1000Hz and a graphic update rate of 30Hz. We also conduct human factor studies to show that the haptic and graphic behavior from our model is realistic. Our experiments verify that our model provides a realistic haptic and graphic feeling to users in real-time.

• International journal of advanced smart convergence
• /
• v.8 no.4
• /
• pp.68-74
• /
• 2019

Along with the deeper architecture in the deep learning approaches, the need for the data becomes very big. In the real problem, to get huge data in some disciplines is very costly. Therefore, learning on limited data in the recent years turns to be a very appealing area. Meta-learning offers a new perspective to learn a model with this limitation. A state-of-the-art model that is made using a meta-learning framework, Meta-SGD, is proposed with a key idea of learning a hyperparameter or a learning rate of the fast adaptation stage in the outer update. However, this learning rate usually is set to be very small. In consequence, the objective function of SGD will give a little improvement to our weight parameters. In other words, the prior is being a key value of getting a good adaptation. As a goal of meta-learning approaches, learning using a single gradient step in the inner update may lead to a bad performance. Especially if the prior that we use is far from the expected one, or it works in the opposite way that it is very effective to adapt the model. By this reason, we propose to add a weight term to decrease, or increase in some conditions, the effect of this prior. The experiment on few-shot learning shows that emphasizing or weakening the prior can give better performance than using its original value.