• Journal of the Korea Computer Graphics Society
• /
• v.25 no.5
• /
• pp.11-20
• /
• 2019

With the development of deep learning technology, research is being actively carried out on user-friendly interfaces that are suitable for use in virtual reality or augmented reality applications. To support the interface using the user's hands, this paper proposes a deep learning-based fingertip detection method to enable the tracking of fingertip coordinates to select virtual objects, or to write or draw in the air. After cutting the approximate part of the corresponding fingertip object from the input image with the Grad-CAM, and perform the convolution neural network with Atrous Convolution for the cut image to detect fingertip location. This method is simpler and easier to implement than existing object detection algorithms without requiring a pre-processing for annotating objects. To verify this method we implemented an air writing application and showed that the recognition rate of 81% and the speed of 76 ms were able to write smoothly without delay in the air, making it possible to utilize the application in real time.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.4
• /
• pp.418-428
• /
• 2002

In this paper, we propose a motion-adaptive de-interlacing method using motion compensated interpolation. In a conventional motion compensated method, a simple pre-filter such as line averaging is applied to interpolate missing lines before the motion estimation. However, this method causes interpolation error because of inaccurate motion estimation and compensation. In the proposed method, EBMF(Edge Based Median Filter) as a pre-filter is applied, and new matching method, which uses two same-parity fields and opposite-parity field as references, is proposed. For further improvement, motion correction filter is proposed to reduce the interpolation error caused by incorrect motion. Simulation results show that the proposed method provides better performance than existing methods.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.5
• /
• pp.1147-1159
• /
• 1996

In this paper. we develop an efficient algorithm which vectorize the table input for mixed document recognition system. It is necessary to separate character and line for recognizing the character in the table. For recognizing table, we have to recognize the character which is blocked by table line and develop the efficient rectorization method for table line. For vectorizing table, we develop several methods. The first method is to extract table line part using 8-dircction chaincodes. The second method is to extract horizontal and vertical lines using histogram of lines. The third one is to extract diagonal lines of table by using the cross points of horizontal and verticallines. Finally we also develop the table vectorization method which finds the regularity characteristics of horizontal and vertical lines composing table, In the paper, we sugest a regularity method for efficient table vectorization.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.5C
• /
• pp.618-626
• /
• 2004

In this paper, we describe a robot endeffector recognition system based on a Modular Neural Networks (MNN). The proposed recognition system can be used for vision system which track a given object using a sequence of images from a camera unit. The main objective to achieve with the designed MNN is to precisely recognize the given robot endeffector and to minimize the processing time. Since the robot endeffector can be viewed in many different shapes in 3- D space, a MNN structure, which contains a set of feedforwared neural networks, can be more attractive in recognizing the given object. Each single neural network learns the endeffector with a cluster of training patterns. The training MNN patterns for a neural network share the similar characteristics so that they can be easily trained. The trained UM is les s sensitive to noise and it shows the better performance in recognizing the endeffector. The recognition rate of MNN is enhanced by 14％ over the single neural network. A vision system with the MNN can precisely recognize the endeffector and place it at the center of a display for a remote operator.

• Journal of Korea Multimedia Society
• /
• v.7 no.1
• /
• pp.35-43
• /
• 2004

In this paper, we proposed a vehicle recognition and electronic toll collection system with detection and classification of vehicle identification mark and emblem as well as recognition of vehicle license plate to unman toll fee collection system or incoming/outcoming vehicles to an institution. In the proposed algorithm, we first process pre-processing step such as noise reduction and thinning from the rear side input image of vehicle and detect vehicle mark, emblem and license plate region using intensity variation informations, template masking and labeling operation. And then, we classify the detected vehicle features regions into vehicle mark and emblem as well as recognize characters and numbers of vehicle license plate using hybrid and seven segment pattern vector. To show the efficiency of the proposed algorithm, we tested it on real vehicle images of implemented vehicle recognition system in highway toll gate and found that the proposed method shows good feature detection/classification performance regardless of irregular environment conditions as well as noise, size, and location of vehicles. And also, the proposed algorithm may be utilized for catching criminal vehicles, unmanned toll collection system, and unmanned checking incoming/outcoming vehicles to an institution.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.8
• /
• pp.1867-1873
• /
• 2015

Edge detection is one of image processing techniques applied for a variety of purposes in a number of areas and it is used as a necessary pretreatment process in most applications. In the conventional edge detection methods, there are Sobel, Prewitt, Roberts and LoG, etc using a fixed weights mask. Since conventional edge detection methods apply the images to the fixed weights mask, the edge detection characteristics appear somewhat insufficient. Therefore in this study, an algorithm for detecting the edge is proposed by applying the cross mask based on the center pixel and up, down, left and right mask based on the surrounding pixels of center pixel in order to solve these problems. And in order to assess the performance of proposed algorithm, it was compared with a conventional Sobel, Roberts, Prewitt and LoG edge detection methods.

• Journal of Korean Society for Geospatial Information Science
• /
• v.13 no.1 s.31
• /
• pp.19-25
• /
• 2005

Nowadays, it is possible to construct the DEMs of urban area effectively and economically by LiDAR system. But the data from LiDAR system has form of DSM which is included various objects as trees and buildings. So the preprocess is necessary to extract the DEMs from LiDAR DSMs for particular purpose as effects analysis of man-made objects for flood prediction. As this study is for extracting DEM from LiDAR DSM of urban area, we detected the edges of various objects using edge detecting algorithm of image process. And, we tried mean value filtering, median value filtering and minimum value filtering or detected edges instead of interpolation method which is used in the previous study and could be modified the source data. it could minimize the modification of source data, and the extracting process of DEMs from DSMs could be simplified and automated.

• Journal of KIISE:Software and Applications
• /
• v.34 no.1
• /
• pp.59-69
• /
• 2007

In the video with a various environment, background modeling is important for extraction and recognition the moving object. For this object recognition, many methods of the background modeling are proposed in a process of preprocess. Among these there is a Kumar method which represents the Queue-based background modeling. Because this has a fixed period of updating examination of the frame, there is a limit for various system. This paper use a background modeling based on the queue. We propose the method that major parameters are decided as adaptive by background model. They are the queue size of the sliding window, the sire of grouping by the brightness of the visual and the period of updating examination of the frame. In order to determine the factors, in every process, RCO (Ratio of Correct Object), REO (Ratio of Error Object) and UR (Update Ratio) are considered to be the standard of evaluation. The proposed method can improve the existing techniques of the background modeling which is unfit for the real-time processing and recognize the object more efficient.

• 한국지구물리탐사학회:학술대회논문집
• /
• 2000.09a
• /
• pp.158-172
• /
• 2000

In order to visualize mine cavities, 3D seismic reflection data were acquired at the ground subsidence sites near Nampung coal mine area, Neukguri, Dogye, Samchuck. Full range 3D array with complete range of azimuths on the bins was considered in the data acquisition design. Because of poor S/N data, we estimated the stacking velocities by CVS method, and we estimated the shot and receiver statics on the shot and receiver stack data. We could confirm that features of ground collapse that were expected from the subsidence. In order to visualize the cavities, we need to apply more sophisticated processing schemes, such as velocity analysis, residual statics correction.

• Journal of Digital Convergence
• /
• v.14 no.8
• /
• pp.233-243
• /
• 2016

Due to speech recognition problems in noisy environment, Audio Visual Speech Recognition (AVSR) system, which combines speech information and visual information, has been proposed since the mid-1990s,. and lip reading have played significant role in the AVSR System. This study aims to enhance recognition rate of utterance word using only lip shape detection for efficient AVSR system. After preprocessing for lip region detection, Convolution Neural Network (CNN) techniques are applied for utterance period detection and lip shape feature vector extraction, and Hidden Markov Models (HMMs) are then used for the recognition. As a result, the utterance period detection results show 91% of success rates, which are higher performance than general threshold methods. In the lip reading recognition, while user-dependent experiment records 88.5%, user-independent experiment shows 80.2% of recognition rates, which are improved results compared to the previous studies.