• Journal of KIISE:Software and Applications
• /
• v.34 no.7
• /
• pp.585-594
• /
• 2007

The back propagation algorithm took a long time to learn the input patterns and was difficult to train the additional or repeated learning patterns. So Aleksander proposed the binary neural network which could overcome the disadvantages of BP Network. But it had the limitation of repeated learning and was impossible to extract a generalized pattern. In this paper, we proposed a dynamic 3 dimensional Neuro System which was consisted of a learning network which was based on weightless neural network and a feedback module which could accumulate the characteristic. The proposed system was enable to train additional and repeated patterns. Also it could be produced a generalized pattern by putting a proper threshold into each learning-net's discriminator which was resulted from learning procedures. And then we reused the generalized pattern to elevate the recognition rate. In the last processing step to decide right category, we used maximum response detector. We experimented using the MNIST database of NIST and got 99.3% of right recognition rate for training data.

• Proceedings of the Korean Geotechical Society Conference
• /
• 2002.03a
• /
• pp.651-656
• /
• 2002

In general, the most important problem in slope stability analysis is that there is no definite way to describe the natural three-dimensional Joint network. Therefore, the many approaches were tried to anlayze the slope stability. Numerical modeling approach is one of the branch to resolve the complexity of natural system. UDEC, FLAC, and SWEDGE are widely used commercial code for the purpose on stability analysis. For the purpose on the more appropriate application of these kind of code, however, three-dimensional distribution of joint network must be identified in more explicit way. Remaining problem is to definitely describe the three dimensional network of joint and bedding, but it is almost impossible in practical sense. Three dimensional joint generation method with random number generation and the results of generation to UDEC have been applied to settle the refered problems in field site. However, this approach also has a important problem, and it is that joint network is generated only once. This problem lead to the limitation on the application to field case, in practical sense. To get rid of this limitation, Monte Carlo Simulation is proposed in this study 1) statistical analysis of input values and definition of the applied system with statistical parameter, 2) instead of the consideration of generated network as a real system, generated system is just taken as one reliable system, 3) present the design parameters, through the statistical analysis of ouput values Results of this study are not only the probability of failure, but also area of failure block, shear strength, normal strength and failure pattern, and all of these results are described in statistical parameters. The results of this study, shear strength, failure area, pattern etc, can provide the direct basement on the design, cutoff angle, support pattern, support strength and etc.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.1
• /
• pp.55-62
• /
• 1996

MRNS network is a general algebraic structure of Hypercube network which has recently drawn considerable attention to supercomputing and message-passing communication. In this paper, we investigate the routing of a message in an n- dimensional MRNS network that is a key to the performance of this network. On the n-dimensional MRNS network we would like to transmit packets from a source node to a destination node simultaneously along a fixed number of paths, where the superscript packet will traverse along the superscript path. In order for all packets to arrive at the destination node quickly and securely, the ith path must be node-disjoint from all other paths. By investigating the conditions of node-disjoint paths, we will employ the special matrices called as the Hamiltonian Circuit Latin Square(HCLS) described in 〔1〕to construct a set of node-disjoint paths and suggest a linear-time parallel routing algorithm for the MRNS network.

• The Transactions of the Korean Institute of Electrical Engineers B
• /
• v.51 no.5
• /
• pp.234-241
• /
• 2002

An optical pickup actuator is an objective-lens-moving mechanism that provides a means to follow the disk displacement accurately(1). In this paper, a slim type optical pickup actuator for Notebook PCs is analyzed and designed to improve the driving sensitivity A three dimensional equivalent magnetic circuit network method (3D-EMCN method) is proposed for an analysis method which provides better characteristics in both precision and computation time of analysis comparing with a commercial three-dimensional finite element (3D-FEM) codes. To verify the validity of proposed method, we made a comparison between the analysis results and the experimental ones. We also compared this analysis results with 3D-FEM results. Among the several optimal algorithm, we adopt a niching genetic algorithm, which renders a set of the multiple optimal solutions. RCS (Restricted Competition Selection) niching genetic algorithm is used for optimal design of the actuator's performance. Recently, the pickup actuator needs additional driving structure for radial and tangential tilting motion to obtain better pick-up performance. So we applied the proposed method to the model containing tilting coils.

• Journal of Electrical Engineering and information Science
• /
• v.3 no.3
• /
• pp.342-347
• /
• 1998

An OptiNeural system is developed for optical pattern classification. It is a novel hybrid system which consists of an optical processor and a multilayer neural network. It takes advantages of two dimensional processing capability of an optical processor and nonlinear mapping capability of a neural network. The optical processor with a binary phase only filter is used as a preprocessor for feature extraction and the neural network is used as a decision system through mapping. OptiNeural system is trained for optical pattern classification by use of a simulated annealing algorithm. Its classification performance for grey tone texture patterns is excellent, while a conventional optical system shows poor classification performance.

• Proceedings of the KIEE Conference
• /
• 1990.11a
• /
• pp.460-464
• /
• 1990

Texture provides an important source of information about the local orientation of visible surfaces. An important task that arises in many computer vision systems is the reconstruction of three-dimensional depth information from two-dimensional images. The surface orientation of texel is classified by the Artificial Neural Network. The classification method to recognize the shape of 3D object with artificial neural network requires less developing time comparing to conventional method. The segmentation problem is assumed to be solved. The surface in view is smooth and is covered with repeated texture elements. In this study, 3D shape reconstruct using interpolation method.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.16 no.5
• /
• pp.179-186
• /
• 2021

For stereo matching based on deep learning, the design of network structure is crucial to the calculation of matching cost, and the time-consuming problem of convolutional neural network in image processing also needs to be solved urgently. In this paper, a method of stereo matching using sparse loss volume in parallax dimension is proposed. A sparse 3D loss volume is constructed by using a wide step length translation of the right view feature map, which reduces the video memory and computing resources required by the 3D convolution module by several times. In order to improve the accuracy of the algorithm, the nonlinear up-sampling of the matching loss in the parallax dimension is carried out by using the method of multi-category output, and the training model is combined with two kinds of loss functions. Compared with the benchmark algorithm, the proposed algorithm not only improves the accuracy but also shortens the running time by about 30%.

• Journal of Power System Engineering
• /
• v.20 no.6
• /
• pp.51-57
• /
• 2016

Experiment is carried out to investigate the mixed convective flow in three-dimensional horizontal rectangular channels filled with high viscous fluid. The particle image velocimetry(PIV) with thermo-sensitive liquid crystal tracers is used for visualizing and analysis. Quantitative data of temperature and velocity are obtained by applying the color-image processing to a visualized image, and neural network is applied to the color-to-temperature calibration. In this study, the fluid used is silicon oil(Pr=909), the aspect ratio(channel width to heigh) is 4 and Reynolds number is $2{\times}10^{-2}$. From the present study, we can visualize the quantitative temperature and velocity of mixed convective flow in three-dimensional horizontal rectangular channels simultaneously.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.47-47
• /
• 2000

Most measuring devices are equipped with RS-232 or GPIB interface for communicating data with computers. If the measuring devices can be accessed by a server computer, the valuable information from the devices can be effectively shared with other computers via internet. But, if the measuring devices and the server computer are too far away, it is difficulty to directly connect them by RS232 interface. PSTN(Public Switched Telephone Network) refers to the world's collection of interconnected voice-oriented public telephone networks. Measuring computer system which is equipped with RS232 interface and modem for PSTN can be introduced to overcome the aforementioned distance problem, In this work, an internet based remote monitoring system which utilizes PSTN and VRML for 3-dimensional GUI is proposed.

• Journal of Institute of Control, Robotics and Systems
• /
• v.11 no.4
• /
• pp.290-296
• /
• 2005

This paper presents the implementation of a crack tracking mobile robot. The crack tracking robot is built for tracking cracks on the pavement. To track cracks, crack must be detected by laser and camera sensors. Laser sensor projects laser on the pavement to detect the discontinuity on the surface and the camera captures the image to find the crack position. Then the robot is commanded to follow the crack. To detect crack position correctly, neural network is used to minimize the positional errors of the captured crack position obtained by transformation from 2 dimensional images to 3 dimensional images.