• Korean Journal of Construction Engineering and Management
• /
• v.10 no.2
• /
• pp.145-154
• /
• 2009

For developing an Automation equipment in construction, it is a key issue to develop 3D modeling technology which can be used for automatically recognizing environmental objects. Recently, for the development of "Intelligent Excavating System(IES), a research developing the real-time 3D terrain modeling technology has been implemented from 2006 in Korea and a stereo vision system is selected as the optimum technology. However, as a result of performance tests implemented in various earth moving environment, the 3D images obtained by stereo vision included considerable noise. Therefore, in this study, for getting rid of the noise which is necessarily generated in stereo image matching, the noise elimination algorithm of stereo-vision images for 3D terrain modeling was developed. The consequence of this study is expected to be applicable in developing an automation equipments which are used in field environment.

• Journal of Digital Convergence
• /
• v.14 no.9
• /
• pp.325-329
• /
• 2016

Facial authentication has the limelight because it has less resistance and it is hard to falsify among various biometric identification. The algorithm of facial authentication can bring about huge difference in accuracy and speed by the algorithm construction. Along with face-extracted data by tracing and extracting pupil, the thesis studied algorithm which extracts data to improve error rate and to accurately authenticate face. It detects face by cascade, selects as significant area, divides the facial area into 4 equal parts to save the coordinate of object. Also, to detect pupil from the eye, the binarization is conducted and it detects pupil by Hough conversion. The core coordinate of detected pupil is saved and calculated to conduct facial authentication through data matching. The thesis studied optimized facial authentication algorithm which accurately calculates facial data with pupil trace.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2015.10a
• /
• pp.93-96
• /
• 2015

The aim of the estimation of disparity map is to find the corresponding pixels from similar two or more images. However, it is a difficult problem to get precise and consistent disparity under a variety of real world situations. In other words, the color values of stereo images are easily influenced by radiometric properties such as illumination direction, illumination color, and camera exposure. Therefore, conventional stereo matching methods can have low performances under radiometric conditions. In this paper, we propose an approaching of disparity map estimation that is reliable in controlling various radiometric variations close to the real environment. This method is motivated by following constancy. Even though each other has different radiometric property in stereo images, intensity of pixels of object have general constancy in specific block. Experimental results show that the proposed method has better performances compared to the comparison group under different radiometric conditions between stereo images. Consequentially, the proposed method is able to estimate the disparity map in stable under various radiometric variations.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.6
• /
• pp.29-36
• /
• 2012

In this paper, we suggest a new method of detecting human skin-color regions from three-dimensional static or dynamic stereoscopic images by effectively integrating depth and color features. The suggested method first extracts depth information that represents the distance between a camera and an object from input left and right stereoscopic images through a stereo matching technique. It then performs labeling for pixels with similar depth features and determines the labeled regions having human skin color as actual skin color regions. Our experimental results show that the suggested skin region extraction method outperforms existing skin detection methods in terms of skin-color region extraction accuracy.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2011.10a
• /
• pp.219-222
• /
• 2011

There are number of methods that track the movement of an object or the change of state, such as Kalman filter, particle filter, dynamic clustering, and so on. Amongst these method, dynamic clustering method is an useful way to track cluster across multiple data frames and analyze their trend. In this paper we suggest the similarity-based dynamic clustering method, and verifies it's performance by simulation. Proposed dynamic clustering method is how to determine the same clusters for each continuative frame. The same clusters have similar characteristics across adjacent frames. The change pattern of cluster's characteristics in each time frame is throughly studied. Clusters in each time frames are matched against each others to see their similarity. Mamdani fuzzy model is used to determine similarity based matching algorithm. The proposed algorithm is applied to radar reflectivity data over time domain. We were able to observe time dependent characteristic of the clusters.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37C no.12
• /
• pp.1271-1280
• /
• 2012

In the recent parking assistant systems, finding out the distance to the object behind a car is often performed by the range sensors such as ultrasonic sensors, radars. However, the installation of additional sensors on the used vehicle could be difficult and require extra cost. On the other hand, the motion stereo technique that extracts distance information using only an image sensor was also proposed. However, In the stereo rectification step, the motion stereo requires good features and exacts matching result. In this paper, we propose a fast algorithm that extracts the accurate distance information for the parallel parking situation using the consecutive images that is acquired by a rear-view camera. The proposed algorithm uses the quadrangle transform of the image, the horizontal line integral projection, and the blocking-based correlation measurement. In the experiment with the magna parallel test sequence, the result shows that the line-accurate distance measurement with the image sequence from the rear-view camera is possible.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.5C
• /
• pp.454-463
• /
• 2010

After the efficiency of H.264 video compression has been announced, it replaced MPEG-2 standard in several applications. So transcoding methods of MPEG-2 to H.264 have been studying because there are variety devices and contents followed by MPEG-2. Although H.264 supported various service such as IPTV, DMB, digital broadcasting etc, but users using MPEG-2 devices cannot accessible to them. This paper propose H.264 to MPEG-2 transcoding for users of MPEG-2 devices without displacement H.264. The proposed method predicted a motion vector for MPEG-2 encoder after it extracted from motion vectors of variable blocks in H.264 to improve processing time. Also it predicted a optimal motion vector using modified boundary matching algorithm after grasped a special character for boundary and background of object. The experimental results from proposed method show a considerable reduction in processing time, as much as 68% averagely, with a small objective quality reduction in PSNR.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.21 no.6
• /
• pp.171-181
• /
• 2011

Proposed in this paper is the intelligent video surveillance system to effectively detect multiple loitering objects even that disappear from the out of camera's field of view and later return to a target zone. After the background and foreground are segmented using Gaussian mixture model and shadows are removed, the objects returning to the target zone is recognized using the chromaticity histogram and the duration of loitering is preserved. For more accurate measurement of the loitering behavior, the camera calibration is also applied to map the image plane to the real-world ground. Hence, the loitering behavior can be detected by considering the time duration of the object's existence in the real-world space. The experiment was performed using loitering video and all of the loitering behaviors are accurately detected.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.11
• /
• pp.3991-4010
• /
• 2021

The Question Similarity Measurement of Chinese Crop Diseases and Insect Pests (QSM-CCD&IP) aims to judge the user's tendency to ask questions regarding input problems. The measurement is the basis of the Agricultural Knowledge Question and Answering (Q & A) system, information retrieval, and other tasks. However, the corpus and measurement methods available in this field have some deficiencies. In addition, error propagation may occur when the word boundary features and local context information are ignored when the general method embeds sentences. Hence, these factors make the task challenging. To solve the above problems and tackle the Question Similarity Measurement task in this work, a corpus on Chinese crop diseases and insect pests(CCDIP), which contains 13 categories, was established. Then, taking the CCDIP as the research object, this study proposes a Chinese agricultural text similarity matching model, namely, the AgrCQS. This model is based on mixed information extraction. Specifically, the hybrid embedding layer can enrich character information and improve the recognition ability of the model on the word boundary. The multi-scale local information can be extracted by multi-core convolutional neural network based on multi-weight (MM-CNN). The self-attention mechanism can enhance the fusion ability of the model on global information. In this research, the performance of the AgrCQS on the CCDIP is verified, and three benchmark datasets, namely, AFQMC, LCQMC, and BQ, are used. The accuracy rates are 93.92%, 74.42%, 86.35%, and 83.05%, respectively, which are higher than that of baseline systems without using any external knowledge. Additionally, the proposed method module can be extracted separately and applied to other models, thus providing reference for related research.

• Journal of the Korea Computer Graphics Society
• /
• v.28 no.3
• /
• pp.125-134
• /
• 2022

This paper implements an AR manual for a vehicle that can be used even in the vehicle interior space where it is difficult to apply the augmentation method of AR content, which is mainly used, and applies a deep learning model to improve the augmentation matching between real space and virtual objects. Through deep learning, the logo of the steering wheel is recognized regardless of the position, angle, and inclination, and 3D interior space coordinates are generated based on this, and the virtual button is precisely augmented on the actual vehicle parts. Based on the same learning model, the function to recognize the main warning light symbols of the vehicle is also implemented to increase the functionality and usability as an AR manual for vehicles.