• The Journal of Korea Robotics Society
• /
• v.10 no.4
• /
• pp.223-229
• /
• 2015

This paper proposes a novel method for detection of hand raising poses from images acquired from a single camera attached to a mobile robot that navigates unknown dynamic environments. Due to unconstrained illumination, a high level of variance in human appearances and unpredictable backgrounds, detecting hand raising gestures from an image acquired from a camera attached to a mobile robot is very challenging. The proposed method first detects faces to determine the region of interest (ROI), and in this ROI, we detect hands by using a HOG-based hand detector. By using the color distribution of the face region, we evaluate each candidate in the detected hand region. To deal with cases of failure in face detection, we also use a HOG-based hand raising pose detector. Unlike other hand raising pose detector systems, we evaluate our algorithm with images acquired from the camera and images obtained from the Internet that contain unknown backgrounds and unconstrained illumination. The level of variance in hand raising poses in these images is very high. Our experiment results show that the proposed method robustly detects hand raising poses in complex backgrounds and unknown lighting conditions.

• Proceedings of the Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography Conference
• /
• 2004.11a
• /
• pp.275-279
• /
• 2004

This study was done to detect the topographic and terrain change of the vicinity of the west coast. To make the basic map of the change in topology and terrain, the mosaic images were made using the images from the satellite, which were given the geometric correction based on the GCP (Ground Control Point) and DEM (Digital Elenation Model) data. The accuracy of the images was examined by .empaling them with CCP through 1:25,000's digital map. After that, among the resultant images of the 1970s and 2000s, those of Sihwa, Hwaong and Ansan, the lands reclaimed by drainage were compared to observe the change in the area. From this study, the accuracy of the images of the west coast from satellite could be acquired and the change of the topology and terrain was detected effectively. From the results, it was known that, in case of the land the topological change was not so big due to the development in the reclaimed land or the bare land. In Sihwa, the size of the land was increased 180 $\textrm{km}^2$ and that of the seashore was decreased 110 km. in Hwaong the size was increased 50 $\textrm{km}^2$ and in Ansan the city space was increased 71 $\textrm{km}^2$ due to the formation of the industrial complex.

• Journal of Korea Multimedia Society
• /
• v.25 no.7
• /
• pp.895-902
• /
• 2022

Building segmentation using satellite imagery such as EO (Electro-Optical) and SAR (Synthetic-Aperture Radar) images are widely used due to their various uses. EO images have the advantage of having color information, and they are noise-free. In contrast, SAR images can identify the physical characteristics and geometrical information that the EO image cannot capture. This paper proposes a learning framework for efficient building segmentation that consists of a teacher-student-based privileged knowledge distillation and deformable convolution block. The teacher network utilizes EO and SAR images simultaneously to produce richer features and provide them to the student network, while the student network only uses EO images. To do this, we present objective functions that consist of Kullback-Leibler divergence loss and knowledge distillation loss. Furthermore, we introduce deformable convolution to avoid pixel-level noise and efficiently capture hard samples such as small and thin buildings at the global level. Experimental result shows that our method outperforms other methods and efficiently captures complex samples such as a small or narrow building. Moreover, Since our method can be applied to various methods.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.153-166
• /
• 2022

Identifying fine cracks in steel bridge facilities is a challenging task of structural health monitoring (SHM). This study proposed an end-to-end crack image segmentation framework based on a one-step Convolutional Neural Network (CNN) for pixel-level object recognition with high accuracy. To particularly address the challenges arising from small object detection in complex background, efforts were made in loss function selection aiming at sample imbalance and module modification in order to improve the generalization ability on complicated images. Specifically, loss functions were compared among alternatives including the Binary Cross Entropy (BCE), Focal, Tversky and Dice loss, with the last three specialized for biased sample distribution. Structural modifications with dilated convolution, Spatial Pyramid Pooling (SPP) and Feature Pyramid Network (FPN) were also performed to form a new backbone termed CrackDet. Models of various loss functions and feature extraction modules were trained on crack images and tested on full-scale images collected on steel box girders. The CNN model incorporated the classic U-Net as its backbone, and Dice loss as its loss function achieved the highest mean Intersection-over-Union (mIoU) of 0.7571 on full-scale pictures. In contrast, the best performance on cropped crack images was achieved by integrating CrackDet with Dice loss at a mIoU of 0.7670.

• Journal of the Korean earth science society
• /
• v.28 no.3
• /
• pp.277-288
• /
• 2007

This study analyzed the visual images covering 'weather' and 'weather forecast' in elementary science textbooks from the Syllabus Period to the 7th national curriculum from the socio-semiotic perspective. The results showed that most of the visual images were 'realistic' which were descriptive of real world phenomena. This means that most of the visual images in elementary science text were familiar to students in every curriculum period. The power relationship in communication between images and students was very complex. The visual images in elementary science textbooks include few geometrical and alphanumeric code in every curriculum period. This study provides a new framework to interpret amount of information, functions of information, structures, and social meanings of visual images. It could be also a beginning stage to introduce the socio-semiotic perspective into choosing visual images for next science textbooks.

• Bulletin of the Korean Chemical Society
• /
• v.32 no.12
• /
• pp.4321-4326
• /
• 2011

A new series of heteroleptic cyclometalated iridium(III) complexes has been synthesized and characterized by absorption, emission and cyclic voltammetry studies: $(pqx)_2Ir(acac)$ (1), $(dmpqx)_2Ir(acac)$ (2) and $(dfpqx)_2Ir(acac)$ (3) where pqx=2-phenylquinoxalinate, dmpqx=2-(2,4-dimethoxyphenyl)quinoxalinate, dfpqx=2-(2,4-difluorophenyl) quinoxalinate and acac=acetylacetonate anion. The reaction of excess acetylacetone with ${\mu}$-chloride-bridged dimeric iridium complex, $[(C\^N)_2Ir({\mu}-Cl)]_2$, gives a complex 1 and an unusual hydridoiridium(III) complex, $(pqx)IrH(acac)_2$ (4). The complex 1, 2 and 3 show their emissions in an orangered region (${\lambda}_{PL,max}$ = 583-616 nm), and the emission maxima can be tuned by the change of substituent at phenyl ring of 2-phenylquinoxaline ligand. The phosphorescent line shape indicates that the emissions originate predominantly from $^3MLCT$ states with little admixture of ligand-based $^3({\pi}-{\pi}^*)$ excited states. The structures of complex 3 and 4 are additionally characterized by a single crystal X-ray diffraction method. The complex 3 shows a distorted octahedral geometry around iridium(III) metal ion. A strong trans influence of the phenyl ring is examined. In complex 4, there are two discrete molecules which are mirror images each other at the ratio of 1:1 in an unit cell. We propose that the phosphorescent complex 1, 2 and 3 are possible candidates for the phosphors in OLEDs applications.

• Proceedings of the IEEK Conference
• /
• 2003.07d
• /
• pp.1291-1294
• /
• 2003

This paper presents an approach to extract license plates from parking regulation images which is captured in various photographing direction and complex background. first, we search each row at regular intervals starting from the bottom of a license-plate image, and we set up a rough region for a certain zone in which the sign of intensity vector changes frequently enough and color of license plate is detected enough, assuming it as a candidate location of a license plate. And then, we extract an elaborate area of a license plate by horizontally and vertically projecting vertical edges. Here, tar types of the private and the public, are easily classified according to the color of extracted plates. To evaluate proposed method, we used 200 actual regulation images. As a result, the proposed method showed extraction rate of 96%, which is 9% higher than the previous method using only intensity vector.

• Journal of Biomedical Engineering Research
• /
• v.9 no.2
• /
• pp.185-194
• /
• 1988

The dual-energy technique win used to separate the bone-only and tissue-only images from the conventional chest images. The equivalent thickness of the basic materials are estimated from low and high energy images of a given complex materials using the attenuation coefficient of ma serial componens. We showed that the image quality of dual-energy imaging method can be influenced by the ponlinearity and noise components of system and spectrum distributions The quantitative analysis of Calcium component was performed by dual-energy technique and it is shown that the concentration of the Calcium could be accurately estimated within 5% error range.

• Journal of Korea Multimedia Society
• /
• v.6 no.4
• /
• pp.594-600
• /
• 2003

In this paper, we propose a detection method for facial features in color images with various backgrounds and face poses. To begin with, the proposed method extracts face candidacy region from images with various backgrounds, which have skin-tone color and complex objects, via the color and edge information of face. And then, by using the elliptical shape property of face, we correct a rotation, scale, and tilt of face region caused by various poses of head. Finally, we verify the face using features of face and detect facial features. In our experimental results, it is shown that accuracy of detection is high and the proposed method can be used in pose-invariant face recognition system effectively

• Maxillofacial Plastic and Reconstructive Surgery
• /
• v.19 no.4
• /
• pp.343-350
• /
• 1997

Diagnosis of maxillofacial lesions is very difficult. Recent developments in computed tomography enable the production of three dimnesional images of complex anatomical structures from a series of conventional computed tomographic sections. Methods of three-dimensional analysis of computed tomographic images have recently been described. Mostly, reports have concentrated on applications relative to congenital deformities. In this report, one method of three dimensional reformatting is reviwes. Images formed by this method have solid surface appearance and can be color enhanced and manipulated to isolate anatomic structures of interest. The program allows tissue densitis, volumes, and distances. This report emphasizes maxillofacial applications other than those previously reported in the surgical and radiological literature.