• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4412-4428
• /
• 2018

The purpose of this work is to solve the problem of representing an entire video using Convolutional Neural Network (CNN) features for human action recognition. Recently, due to insufficient GPU memory, it has been difficult to take the whole video as the input of the CNN for end-to-end learning. A typical method is to use sampled video frames as inputs and corresponding labels as supervision. One major issue of this popular approach is that the local samples may not contain the information indicated by the global labels and sufficient motion information. To address this issue, we propose a binary hashing method to enhance the local feature extractors. First, we extract the local features and aggregate them into global features using maximum/minimum pooling. Second, we use the binary hashing method to capture the motion features. Finally, we concatenate the hashing features with global features using different normalization methods to train the classifier. Experimental results on the JHMDB and MPII-Cooking datasets show that, for these new local features, binary hashing mapping on the sparsely sampled features led to significant performance improvements.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.8 no.4
• /
• pp.239-247
• /
• 2008

In a resource-limited wireless communication environment, various approaches to meet the ever growing application requirements in an efficient and transparent manner, are being researched and developed. Amongst many approaches, cross layer technique is by far one of the significant contributions that has undoubtedly revolutionized the way conventional layered architecture is perceived. In this paper, we propose a Universal Cross Layer Framework based on vertical layer architecture. The primary contribution of this paper is the functional architecture of the vertical layer which is primarily responsible for cross layer interaction management and optimization. The second contribution is the use of optimization cycle that comprises awareness parameters collection, mapping, classification and the analysis phases. The third contribution of the paper is the decomposition of the parameters into local and global network perspective for opportunistic optimization. Finally, we have shown through simulations how parameters' variations can represent local and global views of the network and how we can set local and global thresholds to perform opportunistic optimization.

• The Journal of Industrial Distribution & Business
• /
• v.12 no.2
• /
• pp.7-17
• /
• 2021

Purpose: Previous studies have rarely examined the role of higher education for human capital development in the case of Korea. The purpose of this study is to investigate effects of higher education for international students in globalized environment on job satisfaction, globalization, and hosting country. Research questions include the following: i) how do application and improvement on work, performance, self-confidence, and use of network affect job satisfaction; ii) how do leadership, policy improvement, public diplomacy, social responsibility and global competence affect globalization?; iii) how do affective, cognitive, and behavioral components affect attitude of hosting country? Research design, data, and methodology: Online survey and interviews were conducted. For quantitative research, this study applied factor and structural equational modeling, while for qualitative research, this study applied cognitive mapping and sentimental analysis. Results: This study found that most of proposed effects showed significant. The results showed consistent for quantitative and qualitative researches. Conclusions: The results implied that higher education in global environment plays a key role to enhance job satisfaction, globalization, and attitude toward the hosting country and contributes to foster international relations significantly. The results also implied that human capital development in globalized environment helps improve global network and public diplomacy.

• Journal of Korea Multimedia Society
• /
• v.25 no.9
• /
• pp.1273-1283
• /
• 2022

This paper presents a locally adaptive HDR algorithm using the integral image and MSRCR for LDR images with inadequate exposure. There are two categories in controlling the dynamic range, which are global and local tone mappings. Since the global ones are relatively simple but have some limitations at considering regional characteristics, the local ones are often utilized and MSRCR is a representative method. MSRCR gives moderate results, but it requires lots of computations for multi-scale surround Gaussian functions and produces the Halo effect around the edges. Therefore, in order to resolve these main problems, the proposed algorithm remarkably reduces the computation of the surrounds due to the use of the integral image. And a set of variable-sized windows is adopted to decrease the Halo effect, according to the type of pixel's region. In addition, an offset controlling function is presented, which is mainly affected to the subjective image quality and based on the global input and the desired output means. As the results, the proposed algorithm no more use Gaussian functions and can reduce the computation amount and the Halo effect.

• Dementia and Neurocognitive Disorders
• /
• v.23 no.2
• /
• pp.89-94
• /
• 2024

Background and Purpose: This study aimed to evaluate the brain magnetic resonance imaging (MRI) of patients with acute transient global amnesia (TGA) using volumetric analysis to verify whether the brains of TGA patients have pre-existing structural abnormalities. Methods: We evaluated the brain MRI data from 87 TGA patients and 20 age- and sex- matched control subjects. We included brain MRIs obtained from TGA patients within 72 hours of symptom onset to verify the pre-existence of structural change. For voxel-based morphometric analyses, statistical parametric mapping was employed to analyze the structural differences between patients with TGA and control subjects. Results: TGA patients exhibited significant volume reductions in the bilateral ventral anterior cingulate cortices (corrected p<0.05). Conclusions: TGA patients might have pre-existing structural changes in bilateral ventral anterior cingulate cortices prior to TGA attacks.

• Journal of Positioning, Navigation, and Timing
• /
• v.13 no.4
• /
• pp.431-439
• /
• 2024

The International GNSS Service (IGS) provides real-time satellite orbit, clock, and bias corrections through its Real-Time Service (RTS). In addition, Galileo has offered the Galileo High Accuracy Service (HAS) since January 24, 2023, further improving positioning accuracy for GPS and Galileo satellites. HAS data are available through Galileo's E6 signal and Ntrip, enhancing positioning accuracy for GPS and Galileo satellites. This study compared Galileo HAS corrections data and IGS RTS corrections data obtained from the BKG Ntrip client (BNC) on February 1 and June 23, 2024. The orbit, clock, and bias corrections of each satellite were analyzed, revealing that HAS and IGS RTS correction values followed similar trends for most satellites. Additionally, satellite position and clock values computed from these corrections were compared with SP3 and CLK data for accuracy. To perform positioning using HAS, a Code-PPP algorithm was developed, and positioning accuracy was evaluated for GPS-only, Galileo-only, and multi-constellation modes using both GPS and Galileo. Tropospheric errors were mitigated using the Global Model of Pressure and Temperature (GPT) with the Global Mapping Function (GMF), and ionospheric corrections were applied using the Global Ionospheric Map (GIM). As a result, sub-meter level positioning accuracy was achieved. Among the correction types, IGS RTS corrections provided better accuracy for GPS-only measurements, while Galileo HAS corrections yielded superior accuracy for Galileo-only positioning.

• Tunnel and Underground Space
• /
• v.32 no.3
• /
• pp.231-242
• /
• 2022

In this study, we developed a robot operating system (ROS)-based autonomous driving robot that estimates the robot's position in underground mines and drives and returns through multiple waypoints. Autonomous driving robots utilize SLAM (Simultaneous Localization And Mapping) technology to generate global maps of driving routes in advance. Thereafter, the shape of the wall measured through the LiDAR sensor and the global map are matched, and the data are fused through the AMCL (Adaptive Monte Carlo Localization) technique to correct the robot's position. In addition, it recognizes and avoids obstacles ahead through the LiDAR sensor. Using the developed autonomous driving robot, experiments were conducted on indoor experimental sites that simulated the underground mine site. As a result, it was confirmed that the autonomous driving robot sequentially drives through the multiple waypoints, avoids obstacles, and returns stably.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.24 no.1 s.173
• /
• pp.173-181
• /
• 2000

Depending upon the method of the surface generation and the quality of the boundary curves provided, the resulting surfaces may have global or local irregularities in many cases. Thus it would be necessary for the designer to evaluate the surface quality and to modify the surface. This capability is very important because the defect of the surface requires the rework of the dies that causes a big loss in cost and delivery time. To simulate the reflection line test in the actual production line, a faster generation algorithm is presented. Among., various surface interrogation methods using reflection lines, Blinn-Newell type of reflection mapping is applied to generate reflection lines on the trimmed NURBS surface. The generation of reflection lines is formulated as a surface-plane intersection problem, and solved by surface-contouring techniques. In addition, a modified reflection map is proposed to eliminate the discontinuity of reflection lines due to the configuration of the reflection map. A fast reflection line algorithm is developed utilizing an efficient traced contouring technique, and proved to be well suited for real-time quality-assessment task.

• Journal of Broadcast Engineering
• /
• v.16 no.6
• /
• pp.1047-1057
• /
• 2011

In this paper, we propose a novel tone mapping method that implements histogram modification framework on two local regions that are classified using K-means clustering algorithm. In addition, we propose automatic parameter tuning method for histogram modification. The proposed method enhances local details better than the global histogram method. Moreover, the proposed method is fully automatic in the sense that it does not require intervention from human to tune parameters that are involved for computing tone mapping functions. In simulations and experimental studies, the proposed method showed better performance than existing histogram modification method.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.29 no.2
• /
• pp.157-164
• /
• 2011

As high-resolution satellite images have enabled large scale topographic mapping and monitoring on global scale with short revisit time, agile sensor orientation, and large swath width, many countries make effort to secure the satellite image information. In Korea, KOMPSAT-2 (KOrea Multi-Purpose SATellite-2) was launched in July 28 2006 with high specification. These satellites have stereo image acquisition capability for 3D mapping and monitoring. To efficiently handle stereo images such as stereo display and monitoring, the accurate epipolar image generation process is prerequisite. However, the process was highly limited due to complexity in epipolar geometry of pushbroom sensor. Recently, the piecewise approach to generate epipolar images using RPC was developed and tested for in-track IKONOS stereo images. In this paper, the piecewise approach was tested for KOMPSAT-2 across-track stereo images to see how accurately KOMPSAT-2 epipolar images can be generated for 3D geospatial applications. In the experiment, two across-track stereo sets from three KOMPSAT-2 images of different dates were tested using RPC as the sensor model. The test results showed that one-pixel level of y-parallax was achieved for manually measured tie points.