• Journal of Broadcast Engineering
• /
• v.19 no.3
• /
• pp.396-404
• /
• 2014

The field of searching clothing, which is very difficult due to the nature of the informal sector, has been in an effort to reduce the recognition error and computational complexity. However, there is no concrete examples of the whole progress of learning and recognizing for cloth, and the related technologies are still showing many limitations. In this paper, the whole process including identifying both the person and cloth in an image and analyzing both its color and texture pattern is specifically shown for classification. Especially, deformable search descriptor, LBPROT_35 is proposed for identifying the pattern of clothing. The proposed method is scale and rotation invariant, so we can obtain even higher detection rate even though the scale and angle of the image changes. In addition, the color classifier with the color space quantization is proposed not to loose color similarity. In simulation, we build database by training a total of 810 images from the clothing images on the internet, and test some of them. As a result, the proposed method shows a good performance as it has 94.4% matching rate while the former Dense-SIFT method has 63.9%.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.7 no.5
• /
• pp.1033-1039
• /
• 2012

In this paper, a simple and robust algorithm is proposed for detecting each steel plate from a image which contains several steel plates. Steel plate is characterized by line edge, so line detection is a fundamental task for analyzing and understanding of steel plate images. To detect the line edge, the proposed algorithm uses the small eigenvalue analysis. The proposed approach scans an input edge image from the top left corner to the bottom right corner with a moving mask. A covariance matrix of a set of edge pixels over a connected region within the mask is determined and then the statistical and geometrical properties of the small eigenvalue of the matrix are explored for the purpose of straight line detection. Using the detected line edges, each plate is determined based on the directional information and the distance information of the line edges. The results of the experiments emphasize that the proposed algorithm detects each steel plate from a image effectively.

• Journal of Convergence for Information Technology
• /
• v.11 no.10
• /
• pp.20-26
• /
• 2021

In this paper, we studied the rotation angle estimation methods required for image alignment in an image recognition environment. In particular, a rotation angle estimation method applicable to a low specification embedded-based environment was proposed and compared with the existing method using complex moment. The proposed method estimates the rotation angle through similarity mathcing of the 1D projection profile along the radial axis after converting an image into polar coordinates. In addition, it is also possible to select a method of using vector sum of the projection profile, which more simplifies the calculation. Through experiments conducted on binary pattern images and gray-scale images, it was shown that the estimation error of the proposed method is not significantly different from that of complex moment-based method and requires less computation and system resources. For future expansion, a study on how to match the rotation center in gray-scale images will be needed.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.77-91
• /
• 2022

Various monitoring systems have been implemented in civil infrastructure to ensure structural safety and integrity. In long-term monitoring, these systems generate a large amount of data, where anomalies are not unusual and can pose unique challenges for structural health monitoring applications, such as system identification and damage detection. Therefore, developing efficient techniques is quite essential to recognize the anomalies in monitoring data. In this study, several machine learning techniques are explored and implemented to detect and classify various types of data anomalies. A field dataset, which consists of one month long acceleration data obtained from a long-span cable-stayed bridge in China, is employed to examine the machine learning techniques for automated data anomaly detection. These techniques include the statistic-based pattern recognition network, spectrogram-based convolutional neural network, image-based time history convolutional neural network, image-based time-frequency hybrid convolution neural network (GoogLeNet), and proposed ensemble neural network model. The ensemble model deliberately combines different machine learning models to enhance anomaly classification performance. The results show that all these techniques can successfully detect and classify six types of data anomalies (i.e., missing, minor, outlier, square, trend, drift). Moreover, both image-based time history convolutional neural network and GoogLeNet are further investigated for the capability of autonomous online anomaly classification and found to effectively classify anomalies with decent performance. As seen in comparison with accuracy, the proposed ensemble neural network model outperforms the other three machine learning techniques. This study also evaluates the proposed ensemble neural network model to a blind test dataset. As found in the results, this ensemble model is effective for data anomaly detection and applicable for the signal characteristics changing over time.

• Journal of the Korea Society of Computer and Information
• /
• v.14 no.9
• /
• pp.11-19
• /
• 2009

A RAM-based Neural Net(RBNN) which has multi-discriminators is more effective than RBNN with a discriminator. Experience Sensitive Cumulative Neural Network and 3-D Neuro System(3DNS) that accumulate the features point improved the performance of BNN, which were enabled to train additional and repeated patterns and extract a generalized pattern. In recognition process of Neural Net with multi-discriminator, the selection of class was decided by the value of MRD which calculates the accumulated sum of each class. But they had a saturation problem of its memory cells caused by learning volume increment. Therefore, the decision of MRD has a low performance because recognition rate is decreased by saturation. In this paper, we propose the method which improve the MRD ability. The method consists of the optimum MRD and the matching ratio prototype to generalized image, the cumulative filter ratio, the gap of prototype response MRD. We experimented the performance using NIST database of NIST without preprocessor, and compared this model with 3DNS. The proposed MRD method has more performance of recognition rate and more stable system for distortion of input pattern than 3DNS.

• Journal of Intelligence and Information Systems
• /
• v.18 no.1
• /
• pp.109-123
• /
• 2012

Many kinds of the education systems are provided to students. Many kinds of the contents like School subjects, license, job training education and so on are provided through many kinds of the media like text, image, video and so on. Students will apply the knowledge they learnt and will use it when they learn other things. In the existing education system, there have been many situations that the education system isn't really helpful to the students because too hard contents are transferred to them or because too easy contents are transferred to them and they learn the contents they already know again. To solve this phenomenon, a method that transfers the most proper lecture contents to the students is suggested in the thesis. Because the difficulty is relative, the contents A can be easier than the contents B to a group of the students and the contents B can be easier than the contents A to another group of the students. Therefore, it is not easy to measure the difficulty of the lecture contents. A method considering this phenomenon to transfer the proper lecture contents is suggested in the thesis. The whole lecture contents are divided into many lecture modules. The students solve the pattern recognition questions, a kind of the prior test questions, before studying the lecture contents and the system selects and provides the most proper lecture module among many lecture modules to the students according to the score about the questions. When the system selects the lecture module and transfer it to the student, the students' answer and the difficulty of the lecture modules are considered. In the existing education system, 1 kind of the content is transferred to various students. If the same lecture contents is transferred to various students, the contents will not be transferred efficiently. The system selects the proper contents using the students' pattern recognition answers. The pattern recognition question is a kind of the prior test question that is developed on the basis of the lecture module and used to recognize whether the student knows the contents of the lecture module. Because the difficulty of the lecture module reflects the all scores of the students' answers, whenever a student submits the answer, the difficulty is changed. The suggested system measures the relative knowledge of the students using the answers and designates the difficulty. The improvement of the suggested method is only applied when the order of the lecture contents has nothing to do with the progress of the lecture. If the contents of the unit 1 should be studied before studying the contents of the unit 2, the suggested method is not applied. The suggested method is introduced on the basis of the subject "English grammar", subjects that the order is not important, in the thesis. If the suggested method is applied properly to the education environment, the students who don't know enough basic knowledge will learn the basic contents well and prepare the basis to learn the harder lecture contents. The students who already know the lecture contents will not study those again and save more time to learn more various lecture contents. Many improvement effects like these and so on will be provided to the education environment. If the suggested method that is introduced on the basis of the subject "English grammar" is applied to the various education systems like primary education, secondary education, job education and so on, more improvement effects will be provided. The direction to realize these things is suggested in the thesis. The suggested method is realized with the MySQL database and Java, JSP program. It will be very good if the suggested method is researched developmentally and become helpful to the development of the Korea education.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.52 no.4
• /
• pp.156-163
• /
• 2015

In this paper, we propose an active water level and distance measurement algorithm using a light beam pattern. On behalf of conventional water level gauge types of pressure, float-well, ultrasonic, radar, and others, recently, extensive research for video analysis based water level measurement methods is gradually increasing as an importance of accurate measurement, monitoring convenience, and much more has been emphasized. By turning a reference light beam pattern on bridge or embankment actively, we suggest a new approach that analyzes and processes the projected light beam pattern image obtained from camera device, measures automatically water level and distance between a camera and a bridge or a levee. As contrasted with conventional methods that passively have to analyze captured video information for recognition of a watermark attached on a bridge or specific marker, we actively use the reference light beam pattern suited to the installed bridge environment. So, our method offers a robust water level measurement. The reasons are as follows. At first, our algorithm is effective against unfavorable visual field, pollution or damage of watermark, and so on, and in the next, this is possible to monitor in real-time the portable-based local situation by day and night. Furthermore, our method is not need additional floodlight. Tests are simulated under indoor environment conditions from distance measurement over 0.4-1.4m and height measurement over 13.5-32.5cm.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.11
• /
• pp.116-123
• /
• 1994

A novel opto-digital pattern recognition method which has shift, rotation, and scale invariant properties is proposed for recognizing two dimensional images having straight lines. The algorithm is composed of three stages. In the first stage the line features of the image are extracted. The second stage imposes the shift, rotation, and scale invariant properties on the extracted features through normalizing procedure. The required normalizing equations are analytically explained. In the last stage, the artificial feedforward neural network is trained with the extracted features. In order to evaluated the proposed algorithm, nine different edge enhnaced binary images composed of straight lines are tested. Thus the proposed algorithm can recognize the patterns event though they are shifted, rotated, and scaled.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.1
• /
• pp.142-152
• /
• 2016

GPS is not able to be used for indoor positioning and currently most of techniques emerging to overcome the limit of GPS utilize private wireless networks. However, these methods require high costs for installation and maintenance, and they are inappropriate to be used in the place where precise positioning is needed as in indoor parking lots. This paper proposes a vehicular indoor positioning method based on QR-code recognition. The method gets an absolute coordinate through QR-code scanning, and obtain the location (an relative coordinate) of a black-box camera using the tilt and roll angle correction through affine transformation, scale transformation, and trigonometric function. Using these information of an absolute coordinate and an relative one, the precise position of a car is estimated. As a result, average error of 13.79cm is achieved and it corresponds to just 27.6% error rate in contrast to 50cm error of the recent technique based on wireless networks.

• International Journal of Contents
• /
• v.5 no.3
• /
• pp.50-56
• /
• 2009

Internet is being used for several activities by a great range of users. These activities include communication, e-commerce, education, and entertainment. Users are required to register regarding website in order to enroll web activities. However, registration can be done by automated hacking software. That software make false enrollments which occupy the resources of the website by reducing the performance and efficiency of servers, even stop the entire web service. It is crucial for the websites to have a system which has the capability of differing human users and computer programs in reading images of text. Completely Automated Public Turing Test to Tell Computers and Human Apart (CAPTCHA) is such a defense system against Optical Character Recognition (OCR) software. OCR can be defined as software which work for defeating CAPTCHA images and make countless number of registrations on the websites. This study proposes a new CAPTCHA context that is Korean CAPTCHA by means of the method which is splitting CAPTCHA images into several parts with random rotation values, and drawing random lines on a grid background by using Korean characters only. Lines are in the same color with the CAPTCHA text and they provide a distortion of image with grid background. Experimental results show that Korean CAPTCHA is a more secure and effective CAPTCHA type for Korean users rather than current CAPTCHA types due to the structure of Korean letters and the algorithm we are using: rotation and splitting. In this paper, the algorithm of our method is introduced in detail.