• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.4
• /
• pp.648-661
• /
• 1997

This paper presents a new method for post-processing of coded images based upon the low-pass filtering followed by the projection onto the NQCS (narrow quantization constraint set). We also investigate how the proposed method works on JPEG-coded real images. The starting point of the QCS-based post-processing techniques is the centroid of the QCS, where the original image belongs. The low-pass filtering followed by the projection onto the QCS makes the images lie on the boundary of the QCS. It is likely that, however, the original image is inside the QCS. Hence projection onto the NQCS gives a lower MSE (mean square error) than does the projection onto the QCS. Simulation results show that setting the narrowing coefficients of the NQCS to be 0.2 yields the best performance in most cases. Even though the JPEG-coded image is low-pass filtered and projected onto the NQCS repeatedly, there is no guarantee that the resultant image has a lower MSE and goes closer to the original image. Thus only one iteration is sufficient for the post-processing of the coded images. This is interesting because the main drawback of the iterative post-processing techniques is the heavy computational burden. The single iteration method reduces the computational burden and gives us an easy way to implement the real time VLSI post-processor.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.12
• /
• pp.2924-2932
• /
• 2015

Image binarization is a process to divide the image into objects and backgrounds, widely applied to the fields of image analysis and its recognition. In the existing method of binarization, there is some uncertainty when there is insufficient brightness gap between objects and backgrounds in setting threshold. The method of fuzzy binarization has improved the features of objects efficiently. However, since this method sets ${\alpha}$-cut value statically, there remain some problems that important features of objects can be lost during binarization. Therefore, in this paper, we propose a binarization method which does not set ${\alpha}$-cut value statically. The proposed method uses fuzzy membership functions calculated by thresholds of mean, iterative, and Otsu binarization. Experiment results show the proposed method binaries various images with less loss than the existing methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.1
• /
• pp.81-98
• /
• 2013

Since Probabilistic Latent Semantic Analysis (PLSA) and Latent Dirichlet Allocation (LDA) were introduced, many revised or extended topic models have appeared. Due to the intractable likelihood of these models, training any topic model requires to use some approximation algorithm such as variational approximation, Laplace approximation, or Markov chain Monte Carlo (MCMC). Although these approximation algorithms perform well, training a topic model is still computationally expensive given the large amount of data it requires. In this paper, we propose a new method, called non-simultaneous sampling deactivation, for efficient approximation of parameters in a topic model. While each random variable is normally sampled or obtained by a single predefined burn-in period in the traditional approximation algorithms, our new method is based on the observation that the random variable nodes in one topic model have all different periods of convergence. During the iterative approximation process, the proposed method allows each random variable node to be terminated or deactivated when it is converged. Therefore, compared to the traditional approximation ways in which usually every node is deactivated concurrently, the proposed method achieves the inference efficiency in terms of time and memory. We do not propose a new approximation algorithm, but a new process applicable to the existing approximation algorithms. Through experiments, we show the time and memory efficiency of the method, and discuss about the tradeoff between the efficiency of the approximation process and the parameter consistency.

• Journal of Digital Contents Society
• /
• v.18 no.5
• /
• pp.989-994
• /
• 2017

In this paper, we propose a pothole detection system using 2D LiDAR and a pothole detection algorithm. Conventional pothole detection methods can be divided into vibration-based method, 3D reconstruction method, and vision-based method. Proposed pothole detection system uses two inexpensive 2D LiDARs and improves pothole detection performance. Pothole detection algorithm is divided into preprocessing for noise reduction, clustering and line extraction for visualization, and gradient function for pothole decision. By using gradient of distance data function, we check the existence of a pothole and measure the depth and width of the pothole. The pothole detection system is developed using two LiDARs, and the 3D pothole detection performance is shown by detecting a pothole with moving LiDAR system.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.10
• /
• pp.935-943
• /
• 2017

CNN (Convolution neural network), which is used for image classification and speech recognition among neural networks learning based on positive data, has been continuously developed to have a high performance structure to date. There are many difficulties to utilize in an embedded system with limited resources. Therefore, we use GPU (General-Purpose Computing on Graphics Processing Units), which is used for general-purpose operation of GPU to solve the problem because we use pre-learned weights but there are still limitations. Since CNN performs simple and iterative operations, the computation speed varies greatly depending on the thread allocation and utilization method in the Single Instruction Multiple Thread (SIMT) based GPGPU. To solve this problem, there is a thread that needs to be relaxed when performing Convolution and Pooling operations with threads. The remaining threads have increased the operation speed by using the method used in the following feature maps and kernel calculations.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.29 no.9 s.240
• /
• pp.1049-1056
• /
• 2005

A conservative pressure-based finite-volume numerical method has been developed for computing flow and heat transfer by using an unstructured grid system. The method admits arbitrary convex polyhedra. Care is taken in the discretization and solution procedures to avoid formulations that are cell-shape-specific. A collocated variable arrangement formulation is developed, i.e. all dependent variables such as pressure and velocity are stored at cell centers. Gradients required for the evaluation of diffusion fluxes and for second-order-accurate convective operators are found by a novel second-order accurate spatial discretization. Momentum interpolation is used to prevent pressure checkerboarding and the SIMPLE algorithm is used for pressure-velocity coupling. The resulting set of coupled nonlinear algebraic equations is solved by employing a segregated approach, leading to a decoupled set of linear algebraic equations fer each dependent variable, with a sparse diagonally dominant coefficient matrix. These equations are solved by an iterative preconditioned conjugate gradient solver which retains the sparsity of the coefficient matrix, thus achieving a very efficient use of computer resources.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.7 no.3
• /
• pp.21-33
• /
• 1997

In this paper, a method of self-organizing for fuzzy logic controller(FLC) through learning of the proper direction of coritrol is proposed. In case of designing a self-organizing FLC for unknown dynamic plants based on the gradient descent method, it is difficult to identify the desirable direction of the change of control inpul. in which the error would be decreased. To resolve this problem, we propose a method as fo1lows:at first, assign representative values for the direction of change of error with respect to control input to each partitioned region of the states, and then, learn the fuzzy control rules using the reinforced representative values through iterative trials. 'The proposed self-organizing FLC has simple structure and it is easy to design. The validity of the proposed method is proved by the computer simulation for an inverted pendulum system.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.52 no.4
• /
• pp.198-204
• /
• 2003

In this paper, we presented a new algebraic iterative algorithm for the optimal control of the nonlinear systems. The algorithm is based on two steps. The first step transforms nonlinear optimal control problem into a sequence of linear optimal control problem using the quasilinearization method. In the second step, TPBCP(two point boundary condition problem) is solved by algebraic equations instead of differential equations using the new integral operational matrix of BPF(block pulse functions). The proposed algorithm is simple and efficient in computation for the optimal control of nonlinear systems and is less error value than that by the conventional matrix. In computer simulation, the algorithm was verified through the optimal control design of synchronous machine connected to an infinite bus.

• Structural Engineering and Mechanics
• /
• v.35 no.3
• /
• pp.349-364
• /
• 2010

The impact behaviour and the impact-induced damage in laminated composite cylindrical shell subjected to transverse impact by a foreign object are studied using three-dimensional non-linear transient dynamic finite element formulation. A layered version of 20 noded hexahedral element incorporating geometrical non-linearity is developed based on total Langragian approach. Non-linear system of equations resulting from non-linear strain displacement relation and non-linear contact loading are solved using Newton-Raphson incremental-iterative method. Some example problems of graphite/epoxy cylindrical shell panels are considered with variation of impactor and laminate parameters and influence of geometrical non-linear effect on the impact response and the resulting damage is investigated.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.48 no.6
• /
• pp.781-788
• /
• 1999

In this paper, an expert system is proposed, which can be effectively applied to the large-scale systems with the diversity time constraints, the objectives and the unfixed system structure. The inference scheme of the expert system have the integrated structure composed of the intuitive inference module and logical inference module in order to support effectively the operating constraints of system. The intuitive inference module is designed using the pattern matching or pattern recognition method in order to search a same or similar pattern under the fixed system structure. On the other hand, the logical inference module is designed as the structure with the multiple inference mode based on the heuristic search method in order to determine the optimal or near optimal control strategies satisfing the time constraints for system events under the unfixed system structure, and in order to use as knowledge generator. Here, inference mode consists of the best-first, the local-minimum tree, the breadth-iterative, the limited search width/time method. Finally, the application results for large-scale distribution SCADA system proves that the inference scheme of the expert system is very effective for the large-scale system. The expert system is implemented in C language for the dynamic mamory allocation method, database interface, compatability.