• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.14 no.5
• /
• pp.2430-2436
• /
• 2013

This study suggests a method of detecting the still object, which becomes a cause of danger within the crossroad. The Inverse Perspective Transform was performed in order to make the object size consistent by being inputted the real-time image from CCTV that is installed within the crossroad. It established the detection area in the image with the perspective transform and generated the adaptative background image with the use of the moving information on object. The detection of the stop object was detected the candidate region of the stop object by using the background-image differential method. To grasp the appearance of truth on the detected candidate region, a method is proposed that uses the gradient information on image and EHD(Edge Histogram Descriptor). To examine performance of the suggested algorithm, it experimented by storing the images in the commuting time and the daytime through DVR, which is installed on the cross street. As a result of experiment, it could efficiently detect the stop vehicle within the detection region inside the crossroad. The processing speed is shown in 13~18 frame per second according to the area of the detection region, thereby being judged to likely have no problem about the real-time processing.

• Korean Journal of Remote Sensing
• /
• v.33 no.4
• /
• pp.423-436
• /
• 2017

This paper presents a deep learning-based road segmentation framework from very high-resolution orthophotos. The proposed method uses Deep Convolutional Autoencoders for end-to-end mapping of orthophotos to road segmentations. In addition, a set of post-processing steps were applied to make the model outputs GIS-ready data that could be useful for various applications. The optimization of the model's parameters is explained which was conducted via grid search method. The model was trained and implemented in Keras, a high-level deep learning framework run on top of Tensorflow. The results show that the proposed model with the best-obtained hyperparameters could segment road objects from orthophotos at an average accuracy of 88.5%. The results of optimization revealed that the best optimization algorithm and activation function for the studied task are Stochastic Gradient Descent (SGD) and Exponential Linear Unit (ELU), respectively. In addition, the best numbers of convolutional filters were found to be 8 for the first and second layers and 128 for the third and fourth layers of the proposed network architecture. Moreover, the analysis on the time complexity of the model showed that the model could be trained in 4 hours and 50 minutes on 1024 high-resolution images of size $106{\times}106pixels$, and segment road objects from similar size and resolution images in around 14 minutes. The results show that the deep learning models such as Convolutional Autoencoders could be a best alternative to traditional machine learning models for road segmentation from aerial photographs.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.18 no.1
• /
• pp.47-53
• /
• 1982

An analysis of convection, in a fully developed laminar steady flow through the vertical square duct under the condition of variational symmetric heat flux, is considered. Finite element solution algorithm by Galerkin's method with triangular elements and linear interpolation polynominals for the temperature and velocity profiles are derived for the vertical square duct. The comparison of temperature distribution due to variational symmetric heat flux in the duct were made with available the other data when the condition of peripheral heat flux were uniform and zero. Numerical values for the dimensionless temperatures and Nusselt numbers at selected Rayleigh numbers and pressure gradient parameters were obtained at a few nodal points for the vertical square ducts and effects of corner in the duct were investigated.

• Wind and Structures
• /
• v.26 no.5
• /
• pp.331-341
• /
• 2018

A numerical study on the flow over a square cylinder in the vicinity of a wall is conducted for different Couette-Poiseuille-based non-uniform flow with the non-dimensional pressure gradient P varying from 0 to 5. The non-dimensional gap ratio L (=$H^{\ast}/a^{\ast}$) is changed from 0.1 to 2, where $H^{\ast}$ is gap height between the cylinder and wall, and $a^{\ast}$ is the cylinder width. The governing equations are solved numerically through finite volume method based on SIMPLE algorithm on a staggered grid system. Both P and L have a substantial influence on the flow structure, time-mean drag coefficient ${\bar{C}}_D$, fluctuating (rms) lift coefficient ($C_L{^{\prime}}$), and Strouhal number St. The changes in P and L leads to four distinct flow regimes (I, II, III and IV). Following the flow structure change, the ${\bar{C}}_D$, $C_L{^{\prime}}$, and St all vary greatly with the change in L and/or P. The ${\bar{C}}_D$ and $C_L{^{\prime}}$ both grow with increasing P and/or L. The St increases with P for a given L, being less sensitive to L for a smaller P (< 2) and more sensitive to L for a larger P (> 2). A strong relationship is observed between the flow regimes and the values of ${\bar{C}}_D$, $C_L{^{\prime}}$ and St. An increase in P affects the pressure distribution more on the top surface than on bottom surface while an increase in L does the opposite.

• Journal of KIISE:Software and Applications
• /
• v.36 no.5
• /
• pp.347-352
• /
• 2009

Nonnegative matrix factorization (NMF) is a popular method for multivariate analysis of nonnegative data, the goal of which is decompose a data matrix into a product of two factor matrices with all entries in factor matrices restricted to be nonnegative. NMF was shown to be useful in a task of clustering (especially document clustering). In this paper we present an algorithm for orthogonal nonnegative matrix factorization, where an orthogonality constraint is imposed on the nonnegative decomposition of a term-document matrix. We develop multiplicative updates directly from true gradient on Stiefel manifold, whereas existing algorithms consider additive orthogonality constraints. Experiments on several different document data sets show our orthogonal NMF algorithms perform better in a task of clustering, compared to the standard NMF and an existing orthogonal NMF.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.39A no.3
• /
• pp.149-154
• /
• 2014

The cross-correntropy concept can be expressed with inner products of two different probability density functions constructed by Gaussian-kernel density estimation methods. Blind algorithms based on the maximization of the cross-correntropy (MCC) and a symbol set of randomly generated N samples yield superior learning performance, but have a huge computational complexity in the update process at the aim of weight adjustment based on the MCC. In this paper, a method of reducing the computational complexity of the MCC algorithm that calculates recursively the gradient of the cross-correntropy is proposed. The proposed method has only O(N) operations per iteration while the conventional MCC algorithms that calculate its gradients by a block processing method has $O(N^2)$. In the simulation results, the proposed method shows the same learning performance while reducing its heavy calculation burden significantly.

• Korean Journal of Cognitive Science
• /
• v.9 no.1
• /
• pp.1-11
• /
• 1998

In this paper, the features which are thought to be used by humans based on the psychological experiment of human information processing are applied to character recognition problem. Man will deal with a little large area information as well as pixel by pixel information. Therefore we define the feature that represents a little wide region I information called region feature, and combine the features derived from region feature and pixel by pixel features that have been used by now. The features we used are the result of region feature based preanalysis, mesh with region attributes, cross distance difference and gradient. The training and test data in the experiment are handwritten Korean alphabets, digits and English alphabets, which are trained on neural network using back propagation algorithm and recognition results are 90.27-93.25%, 98.00% and 79.73-85.75%, respectively Experimental results show that the feature we are suggesting in this paper is 1-2% better than UDLRH feature similar in attribute to region feature, and the tendency of misrecognition is more easily acceptable by humans.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.2
• /
• pp.145-152
• /
• 2003

In this paper, assuming that the score of speech utterance is the product of HMM log likelihood and HMM weight, we propose a new method that HMM weights are adapted iteratively like the general MCE training. The proposed method adjusts HMM weights for better performance using delta coefficient defined in terms of misclassification measure. Therefore, the parameter estimation and the Viterbi algorithms of conventional 1:.um can be easily applied to the proposed model by constraining the sum of HMM weights to the number of HMMs in an HMM set. Comparing with the general segmental MCE training approach, computing time decreases by reducing the number of parameters to estimate and avoiding gradient calculation through the optimal state sequence. To evaluate the performance of HMM-based speech recognizer by weighting HMM likelihood, we perform Korean isolated digit recognition experiments. The experimental results show better performance than the MCE algorithm with state weighting.

• Journal of the Korea Society of Computer and Information
• /
• v.3 no.2
• /
• pp.57-67
• /
• 1998

In order to extract the contour of interesting object in the image, Kass suggested the Active Contour Model called "Snakes" The speed of this model is slow and this model is sensitive of initialization. In order to improve these problems, Gunn extracted the accurate contour by using two initializations, and operated to less sensitive of initialization. This method could extract more accurate contour than the existing method, but it had no effect in the speed and it was sensitive of noise. This paper applied to the Energy Minimization Algorithm about only the pixel within the window applying the window of $8{\times}8$ size at each contour point consisting Snakes in order to solve these problems. In order to less sensitive of noise which exists within image, it suggests a method that moves the window to vertical direction for the gradient of each contour point.our point.

• International Journal of Control, Automation, and Systems
• /
• v.1 no.2
• /
• pp.194-202
• /
• 2003

In this paper, we introduce a category of Multi-FNN (Fuzzy-Neural Networks) models, analyze the underlying architectures and propose a comprehensive identification framework. The proposed Multi-FNNs dwell on a concept of fuzzy rule-based FNNs based on HCM clustering and evolutionary fuzzy granulation, and exploit linear inference being treated as a generic inference mechanism. By this nature, this FNN model is geared toward capturing relationships between information granules known as fuzzy sets. The form of the information granules themselves (in particular their distribution and a type of membership function) becomes an important design feature of the FNN model contributing to its structural as well as parametric optimization. The identification environment uses clustering techniques (Hard C - Means, HCM) and exploits genetic optimization as a vehicle of global optimization. The global optimization is augmented by more refined gradient-based learning mechanisms such as standard back-propagation. The HCM algorithm, whose role is to carry out preprocessing of the process data for system modeling, is utilized to determine the structure of Multi-FNNs. The detailed parameters of the Multi-FNN (such as apexes of membership functions, learning rates and momentum coefficients) are adjusted using genetic algorithms. An aggregate performance index with a weighting factor is proposed in order to achieve a sound balance between approximation and generalization (predictive) abilities of the model. To evaluate the performance of the proposed model, two numeric data sets are experimented with. One is the numerical data coming from a description of a certain nonlinear function and the other is NOx emission process data from a gas turbine power plant.