• The Journal of the Korea Contents Association
• /
• v.10 no.5
• /
• pp.416-421
• /
• 2010

The KISTI (Korea Institute of Science and Technology Information) began to produce the Korean human information called Visible Korean and Digital Korean since 2000 because there was no human information in Korea which could represent the physical characteristics of Korean human body. The Visible Korean consists of CT, MR, sectioned and segmented images of Korean human body. We obtained the serially sectioned images by grinding the Korean cadaver in horizontal direction and segmented these images by outlining the inner organs of human. We have produced the sectioned images of Korean male whole body, male head, and female pelvis in2008. The segmentation and 3D reconstruction of these images are now in proceeding. The Digital Korean consists of CT images of about 100 Korean cadavers. These CT images were segmented by individual bone, reconstructed to produce the 3D bone models and the skin surface model was also added. The mechanical properties of individual bones were obtained by measuring the property of individual bone sample. We have distributed these Korean human informations to users in domestic and abroad. About 70 institutes in domestic, and 20 institutes in abroad have used our data in research use and nearly 160 proceedings and articles were published since 2001. We think these human informations have a role of medical information infrastructure that could be used in the field of medical education, biomechanics, virtual reality etc.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.3
• /
• pp.237-246
• /
• 2020

Recently, as high resolution satellites data have been serviced, frequent DSM (Digital Surface Model) generation over urban areas has been possible. In addition, it is possible to detect changes using a high-resolution DSM at building level such that various methods of building change detection using DSM have been studied. In order to detect building changes using DSM, we need to generate a DSM using a stereo satellite image. The change detection method using D-DSM (Differential DSM) uses the elevation difference between two DSMs of different dates. The D-DSM method has difficulty in applying a precise vertical threshold, because between the two DSMs may have elevation errors. In this study, we focus on the urban structure change detection using D-nDSM (Differential nDSM) based on nDSM (Normalized DSM) that expresses only the height of the structures or buildings without terrain elevation. In addition, we attempted to reduce noise using a morphological filtering. Also, in order to improve the roadside buildings extraction precision, we exploited the urban road network extraction from nDSM. Experiments were conducted for high-resolution stereo satellite images of two periods. The experimental results were compared for D-DSM, D-nDSM, and D-nDSM with road extraction methods. The D-DSM method showed the accuracy of about 30% to 55% depending on the vertical threshold and the D-nDSM approaches achieved 59% and 77.9% without and with the morphological filtering, respectively. Finally, the D-nDSM with the road extraction method showed 87.2% of change detection accuracy.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.182-185
• /
• 2005

Satellite cameras are calibrated before launch in detail and in general, but it cannot be guaranteed that the geometry is not changing during launch and caused by thermal influence of the sun in the orbit. Modem satellite imaging systems are based on CCD-line sensors. Because of the required high sampling rate the length of used CCD-lines is limited. For reaching a sufficient swath width, some CCD-lines are combined to a longer virtual CCD-line. The images generated by the individual CCD-lines do overlap slightly and so they can be shifted in x- and y-direction in relation to a chosen reference image just based on tie points. For the alignment and difference in scale, control points are required. The resulting virtual image has only negligible errors in areas with very large difference in height caused by the difference in the location of the projection centers. Color images can be related to the joint panchromatic scenes just based on tie points. Pan-sharpened images may show only small color shifts in very mountainous areas and for moving objects. The direct sensor orientation has to be calibrated based on control points. Discrepancies in horizontal shift can only be separated from attitude discrepancies with a good three-dimensional control point distribution. For such a calibration a program based on geometric reconstruction of the sensor orientation is required. The approximations by 3D-affine transformation or direct linear transformation (DL n cannot be used. These methods do have also disadvantages for standard sensor orientation. The image orientation by geometric reconstruction can be improved by self calibration with additional parameters for the analysis and compensation of remaining systematic effects for example caused by a not linear CCD-line. The determined sensor geometry can be used for the generation? of rational polynomial coefficients, describing the sensor geometry by relations of polynomials of the ground coordinates X, Y and Z.

• Journal of Biomedical Engineering Research
• /
• v.44 no.3
• /
• pp.182-189
• /
• 2023

CT is a medical device that acquires medical images based on Attenuation coefficient of human organs related to X-rays. In addition, using this theory, it can acquire sagittal and coronal planes and 3D images of the human body. Then, CT is essential device for universal diagnostic test. But Exposure of CT scan is so high that it is regulated and managed with special medical equipment. As the special medical equipment, CT must implement quality control. In detail of quality control, Spatial resolution of existing phantom imaging tests, Contrast resolution and clinical image evaluation are qualitative tests. These tests are not objective, so the reliability of the CT undermine trust. Therefore, by applying an artificial intelligence classification model, we wanted to confirm the possibility of quantitative evaluation of the qualitative evaluation part of the phantom test. We used intelligence classification models (VGG19, DenseNet201, EfficientNet B2, inception_resnet_v2, ResNet50V2, and Xception). And the fine-tuning process used for learning was additionally performed. As a result, in all classification models, the accuracy of spatial resolution was 0.9562 or higher, the precision was 0.9535, the recall was 1, the loss value was 0.1774, and the learning time was from a maximum of 14 minutes to a minimum of 8 minutes and 10 seconds. Through the experimental results, it was concluded that the artificial intelligence model can be applied to CT implements quality control in spatial resolution and contrast resolution.

• Clinics in Shoulder and Elbow
• /
• v.21 no.3
• /
• pp.151-157
• /
• 2018

Background: Nonunion of lateral humeral condyle fracture causes cubitus valgus deformity. Although corrective osteotomy or osteosynthesis can be considered, there are controversies regarding its treatment. To evaluate elbow joint biomechanics in non-united lateral humeral condyle fractures, we analyzed the motion of elbow joint and pseudo-joint via in vivo three-dimensional (3D) kinematics, using 3D images obtained by computed tomography (CT) scan. Methods: Eight non-united lateral humeral condyle fractures with cubitus valgus and 8 normal elbows were evaluated in this study. CT scan was performed at 3 different elbow positions (full flexion, $90^{\circ}$ flexion and full extension). With bone surface model, 3D elbow motion was reconstructed. We calculated the axis of rotation in both the normal and non-united joints, as well as the rotational movement of the ulno-humeral joint and pseudo-joint of non-united lateral condyle in 3D space from full extension to full flexion. Results: Ulno-humeral joint moved to the varus on the coronal plane during flexion, $25.45^{\circ}$ in the non-united cubitus valgus group and $-2.03^{\circ}$ in normal group, with statistically significant difference. Moreover, it moved to rotate externally on the axial plane $-26.75^{\circ}$ in the non-united cubitus valgus group and $-3.09^{\circ}$ in the normal group, with statistical significance. Movement of the pseudo-joint of fragment of lateral condyle showed irregular pattern. Conclusions: The non-united cubitus valgus group moved to the varus with external rotation during elbow flexion. The pseudo-joint showed a diverse and irregular motion. In vivo 3D motion analysis for the non-united cubitus valgus could be helpful to evaluate its kinematics.

• Journal of Broadcast Engineering
• /
• v.23 no.3
• /
• pp.351-360
• /
• 2018

Human emotion recognition is a research topic that is receiving continuous attention in computer vision and artificial intelligence domains. This paper proposes a method for classifying human emotions through multiple neural networks based on multi-modal signals which consist of image, landmark, and audio in a wild environment. The proposed method has the following features. First, the learning performance of the image-based network is greatly improved by employing both multi-task learning and semi-supervised learning using the spatio-temporal characteristic of videos. Second, a model for converting 1-dimensional (1D) landmark information of face into two-dimensional (2D) images, is newly proposed, and a CNN-LSTM network based on the model is proposed for better emotion recognition. Third, based on an observation that audio signals are often very effective for specific emotions, we propose an audio deep learning mechanism robust to the specific emotions. Finally, so-called emotion adaptive fusion is applied to enable synergy of multiple networks. The proposed network improves emotion classification performance by appropriately integrating existing supervised learning and semi-supervised learning networks. In the fifth attempt on the given test set in the EmotiW2017 challenge, the proposed method achieved a classification accuracy of 57.12%.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.10 no.2
• /
• pp.133-140
• /
• 2017

In this paper, we has proposed an automatic vehicle monitoring system based on computer vision technology. The real-time display system has displayed a system that can be performed in automatic monitoring and control while meeting the essential requirements of ITS. Another advantage has that for a powerful vehicle tracking, the main obstacle handing system, which has the shadow tracking of moving objects. In order to obtain all kinds of information from the tracked vehicle image, the vehicle must be clearly displayed on the surveillance screen. Over time, it's necessary to precisely control the vehicle, and a three-dimensional model-based approach has been also necessary. In general, each type of vehicle has represented by the skeleton of the object or wire frame model, and the trajectory of the vehicle can be measured with high precision in a 3D-based manner even if the system has not running in real time. In this paper, we has applied on segmentation method to vehicle, background, and shadow. The validity of the low level vehicle control tracker was also detected through speed tracking of the speeding car. In conclusion, we intended to improve the improved tracking method in the tracking control system and to develop the highway monitoring and control system.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.10
• /
• pp.11-17
• /
• 2022

In this paper, a model that can increase the efficiency of work in arranging interior furniture by applying augmented reality technology was studied. In the existing system to which augmented reality is currently applied, there is a problem in that information is limitedly provided depending on the size and nature of the company's product when outputting the image of furniture. To solve this problem, this paper presents an AR labeling algorithm. The AR labeling algorithm extracts feature points from the captured images and builds a database including indoor location information. A method of detecting and learning the location data of furniture in an indoor space was adopted using the CNN technique. Through the learned result, it is confirmed that the error between the indoor location and the location shown by learning can be significantly reduced. In addition, a study was conducted to allow users to easily place desired furniture through augmented reality by receiving detailed information about furniture along with accurate image extraction of furniture. As a result of the study, the accuracy and loss rate of the model were found to be 99% and 0.026, indicating the significance of this study by securing reliability. The results of this study are expected to satisfy consumers' satisfaction and purchase desires by accurately arranging desired furniture indoors through the design and implementation of AR labels.

• Journal of the Korean Society of Visualization
• /
• v.7 no.2
• /
• pp.64-71
• /
• 2010

The flow and noise characteristics of wake behind wind-turbine blades have been investigated experimentally using a two-frame particle image velocimetry (PIV) technique. Experiments were carried out in a POSTECH subsonic large wind-tunnel ($1.8^W{\times}1.5^H{\times}4.3^L\;m^3$) with KBP-750D (3-blade type) wind-turbine model at a freestream velocity of $U_o\;=\;15\;m/s$ and a tip speed ratio $\lambda\;=\;6.14$ (2933 rpm). The wind-turbine blades are connected to an AC servo motor, brake, encoder and torque meter to control the rotational speed and to extract a synchronization signal for PIV measurements. The wake flow was measured at four azimuth angles ($\phi\;=\;0^{\circ}$, $30^{\circ}$, $60^{\circ}$ and $90^{\circ}$) of the wind-turbine blade. The dominant flow structure of the wake is large-scale tip vortices. The turbulent statistics such as turbulent intensity are weakened as the flow goes downstream due to turbulent dissipation. The dominant peak frequency of the noise signal is identical to the rotation frequency of blades. The noise seems to be mainly induced by the tip vortices.

• Journal of Welding and Joining
• /
• v.35 no.3
• /
• pp.35-43
• /
• 2017

A 3D transient heat transfer model is developed by ABAQUS software to study the temperature distribution during friction stir welding process at different rotational speeds. Furthermore, AA 5083-O plates were joined by FSW technique. For this purpose, a universal milling machine was used to perform the welding process and a mechanical vice was used to fix the work pieces in the proper position. The joints were friction stir welded at a constant travel speed 50 mm/min and two rotational speed values; 400 rpm and 630 rpm using two types of tools; cylindrical threaded pin and tapered smooth one. At each welding condition the temperature was measured using infra-red thermal image camera to verify the simulated temperature distribution. The welded joints were visually inspected as well as by macro- and microstructure evolutions. In addition, the welded joints were mechanically tested for hardness and tensile strength. The maximum peak temperature obtained was at higher rotational speed using the threaded tool pin profile. The results showed that the rotational speed affects the peak temperature, defects formation and sizes, and the mechanical properties of friction stir welded joints. Moreover, the threaded tool gives superior mechanical properties than the tapered one at lower rotational speed.