• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.16 no.4
• /
• pp.246-253
• /
• 2016

Infrared (IR) imaging has been researched for various applications such as surveillance. IR radiation has the capability to detect thermal characteristics of objects under low-light conditions. However, automatic segmentation for finding the object of interest would be challenging since the IR detector often provides the low spatial and contrast resolution image without color and texture information. Another hindrance is that the image can be degraded by noise and clutters. This paper proposes multi-level segmentation for extracting regions of interest (ROIs) and objects of interest (OOIs) in the IR scene. Each level of the multi-level segmentation is composed of a k-means clustering algorithm, an expectation-maximization (EM) algorithm, and a decision process. The k-means clustering initializes the parameters of the Gaussian mixture model (GMM), and the EM algorithm estimates those parameters iteratively. During the multi-level segmentation, the area extracted at one level becomes the input to the next level segmentation. Thus, the segmentation is consecutively performed narrowing the area to be processed. The foreground objects are individually extracted from the final ROI windows. In the experiments, the effectiveness of the proposed method is demonstrated using several IR images, in which human subjects are captured at a long distance. The average probability of error is shown to be lower than that obtained from other conventional methods such as Gonzalez, Otsu, k-means, and EM methods.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2007.11a
• /
• pp.754-759
• /
• 2007

This study exploits the data discriminating capability of silhouette statistics, which combines wavelet-based vertical energy threshold technique for the purpose of extracting damage-sensitive features and clustering signals of the same class. This threshold technique allows to first obtain a suitable subset of the extracted or modified features of our data, i.e., good predictor sets should contain features that are strongly correlated to the characteristics of the data without considering the classification method used, although each of these features should be as uncorrelated with each other as possible. The silhouette statistics have been used to assess the quality of clustering by measuring how well an object is assigned to its corresponding cluster. We use this concept for the discriminant power function used in this paper. The simulation results of damage detection in a truss structure show that the approach proposed in this study can be successfully applied for locating both open- and breathing-type damage even in the presence of a considerable amount of process and measurement noise.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.12
• /
• pp.1254-1258
• /
• 2010

In information retrieval, different surface forms of the same object can cause poor performance of systems. In this paper, we propose the method extracting alternative words using translation words as features of each word extracted from parallel corpus, korean/english title pair of patent information. Also, we propose an association word filtering method to remove association words from an alternative word list. Evaluation results show that the proposed method outperforms other alternative word extraction methods.

• Journal of Korea Multimedia Society
• /
• v.4 no.3
• /
• pp.215-221
• /
• 2001

In this paper, we propose a algorithm that apply different weight-sampling values according to the directions of the contour to reduce errors that can arise in extracting the feature of an contoured object. Especially, it 8is designed to have invariant property under the circumstances like the rotation, transition and scaling. The output matrix of feature set is made through the size function of the proposed algorithm, and the euclidean distance between the output matrix and that of the original image is calculated. Experimental result shows that the proposed algorithm reduces the euclidean distance between the original image and the changed image-by 57% in rotation and by 91% in scaling.

• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.8
• /
• pp.309-314
• /
• 2015

One of typical methods for extracting facial features from face images may be snake. Although snake is simple and fast, performance is very much affected by the initial contour and the shape of object to be extracted. In this paper, the enhanced snake is proposed to extract better facial features from 6 lip and mouth images as snake point is added to the midpoint of snake segment. It is shown that RSD of the proposed method is about 2.8% to 5.8% less than that of Greedy snake about 6 test face images. Since lesser RSD is especially obtained for contours with highly concavity, the contour is more accurately extracted.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.11
• /
• pp.5527-5545
• /
• 2019

Through-wall ultra-wide band (UWB) radar has been considered as one of the preferred and non-contact technologies for the targets detection owing to the better time resolution and stronger penetration. The high time resolution is a result of a larger of bandwidth of the employed UWB pulses from the radar system, which is a useful tool to separate multiple targets in complex environment. The article emphasised on human subject localization and detection. Human subject usually can be detected via extracting the weak respiratory signals of human subjects remotely. Meanwhile, the range between the detection object and radar is also acquired from the 2D range-frequency matrix. However, it is a challenging task to extract human respiratory signals owing to the low signal to clutter ratio. To improve the feasibility of human respiratory signals detection, a new method is developed via analysing the standard deviation based kurtosis of the collected pulses, which are modulated by human respiratory movements in slow time. The range between radar and the detection target is estimated using joint time-frequency analysis (JTFA) of the analysed characteristics, which provides a novel preliminary signature for life detection. The breathing rates are obtained using the proposed accumulation method in time and frequency domain, respectively. The proposed method is validated and proved numerically and experimentally.

• The Journal of the Korea Contents Association
• /
• v.6 no.4
• /
• pp.98-108
• /
• 2006

In the past, the Image processing system is independently implemented and has a limit in its application to a degree of simple display. The scope of present image processing system is diversely extended in its application owing to the development of image processing IC chips. In this paper, we implement the image processing system operated independently without PC by converting analogue image signals into digital signals. In the proposed image processing system, we extract the motion parameters from analogue image signals and generate the virtual movement to Simulator and operate Simulator by extracting motion parameters.

• Journal of Korea Multimedia Society
• /
• v.11 no.5
• /
• pp.567-578
• /
• 2008

For analyzing a huge amount of web pages available in the Internet, we need to extract the encoded information in web pages. In this paper, we propose a method to extract and convert web information from web pages into XML documents for multidimensional analysis. For extracting information from web pages, we propose two languages: one for describing web information extraction rules based on the object-oriented model, and another for describing regular expressions of HTML tag patterns to search for target information. For multidimensional analysis on XML documents, we propose a method for constructing an XML warehouse and various XML cubes from it like the way we do for relational data. Finally, we show the validness of our method through the application to US patent web pages.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.2
• /
• pp.157-163
• /
• 2011

This paper proposes a hybrid image stabilizing system which uses both optical image stabilizing system based on EKF (Extended Kalman Filter) and digital image stabilization based on SURF (Speeded Up Robust Feature). Though image information is one of the most efficient data for object recognition, it is susceptible to noise which results from internal vibration as well as external factors. The blurred image obtained by the camera mounted on a robot makes it difficult for the robot to recognize its environment. The proposed system estimates shaking angle through EKF based on the information from inclinometer and gyro sensor to stabilize the image. In addition, extracting the feature points around rotation axis using SURF which is robust to change in scale or rotation enhances processing speed by removing unnecessary operations using Hessian matrix. The experimental results using the proposed hybrid system shows its effectiveness in extended frequency range.

• Journal of the Korean Society for Precision Engineering
• /
• v.10 no.4
• /
• pp.81-93
• /
• 1993

The three dimensional(3D) shape reconstruction from two dimensional(2D) cross-sections can be completed through three main phases : the input compilation, the triangular grid formation, and the smooth surface construction. In the input compilation phase, the cross-sections are analyzed to exctract the input data required for the shape reconstruction. This data includes the number of polygonized contours per cross-section and the vertices defining each polygonized contour. In the triangular grid formation phase, a triangular grid, leading to a polyhedral approximations, is constructed by extracting all the information concerning contour links between two adjacent cross- sections and then performing the appropriate triangulation procedure for each contour link. In the smooth surface construction phase, a smooth composite surface interpolating all vertices on the triangular grid is constructed. Both the smooth surface and the polyhedral approximation can be used as reconstructed models of the object. This paper proposes a new method for reconstructing the geometric model of a 3D objdect from a sequence of planar contours representing 2D cross-sections of the objdect. The method includes the triangular grid formation algorithms for contour closing, one-to-one branching, and one-to-many braanching, and many-to-many branching. The shape reconstruction method has been implemented on a SUN workstation in C.