• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.8
• /
• pp.1013-1018
• /
• 2021

Due to the recent accident caused by the automated vehicle, discussions on the ethical aspects of AI have been actively underway. This paper confirms that AI is inevitably connected to ethical components through the concepts and techniques related to robots-AI, and argues that ethical aspects are built-in, not post facto. Furthermore, this devises a solution to the trolley dilemma that can serve as a clue to ethical problems associated with automated vehicles. Preferentially, that process contains writing Bayesian networks. Next, only important and influential data are left after the pre-processing stage, and crowd-sourcing & extrapolation is used to calculate the exact figures of the networks. Through this process, this argues that humans' subjects are certainly included in implementing algorithms and models and discusses the necessity and direction of engineering liberal arts, especially education of ethics that distinguished from major education to prevent distortions and biases abouts AI systems.

• Atmosphere
• /
• v.31 no.5
• /
• pp.489-510
• /
• 2021

In this paper, we propose an algorithm for detecting convective initiation (CI) using GEO-KOMPSAT-2A/advanced meteorological imager data. The algorithm identifies clouds that are likely to grow into convective clouds with radar reflectivity greater than 35 dBZ within the next two hours. This algorithm is developed using statistical and qualitative analysis of cloud characteristics, such as atmospheric instability, cloud top height, and phase, for convective clouds that occurred on the Korean Peninsula from June to September 2019. The CI algorithm consists of four steps: 1) convective cloud mask, 2) cloud object clustering and tracking, 3) interest field tests, and 4) post-processing tests to remove non-convective objects. Validation, performed using 14 CI events that occurred in the summer of 2020 in Korean Peninsula, shows a total probability of detection of 0.89, false-alarm ratio of 0.46, and mean lead-time of 39 minutes. This algorithm can be useful warnings of rapidly developing convective clouds in future by providing information about CI that is otherwise difficult to predict from radar or a numerical prediction model. This CI information will be provided in short-term forecasts to help predict severe weather events such as localized torrential rainfall and hail.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.7
• /
• pp.2131-2153
• /
• 2022

To achieve accurate detection of tuberculosis (TB) areas in chest radiographs, we design a chest X-ray TB area detection algorithm. The algorithm consists of two stages: the chest X-ray TB classification network (CXTCNet) and the chest X-ray TB area detection network (CXTDNet). CXTCNet is used to judge the presence or absence of TB areas in chest X-ray images, thereby excluding the influence of other lung diseases on the detection of TB areas. It can reduce false positives in the detection network and improve the accuracy of detection results. In CXTCNet, we propose a channel attention mechanism (CAM) module and combine it with DenseNet. This module enables the network to learn more spatial and channel features information about chest X-ray images, thereby improving network performance. CXTDNet is a design based on a sparse object detection algorithm (Sparse R-CNN). A group of fixed learnable proposal boxes and learnable proposal features are using for classification and location. The predictions of the algorithm are output directly without non-maximal suppression post-processing. Furthermore, we use CLAHE to reduce image noise and improve image quality for data preprocessing. Experiments on dataset TBX11K show that the accuracy of the proposed CXTCNet is up to 99.10%, which is better than most current TB classification algorithms. Finally, our proposed chest X-ray TB detection algorithm could achieve AP of 45.35% and AP50 of 74.20%. We also establish a chest X-ray TB dataset with 304 sheets. And experiments on this dataset showed that the accuracy of the diagnosis was comparable to that of radiologists. We hope that our proposed algorithm and established dataset will advance the field of TB detection.

• Atmosphere
• /
• v.31 no.5
• /
• pp.473-488
• /
• 2021

A new physical/statistical diagnostic downscale model has been developed for use to improve near-surface air temperature forecasts. The model includes a series of physical and statistical correction methods that account for un-resolved topographic and land-use effects as well as statistical bias errors in a low-resolution atmospheric model. Operational temperature forecasts of the Local Data Assimilation and Prediction System (LDAPS) were downscaled at 100 m resolution for three months, which were used to validate the model's physical and statistical correction methods and to compare its performance with the forecasts of the Korea Meteorological Administration Post-processing (KMAP) system. The validation results showed positive impacts of the un-resolved topographic and urban effects (topographic height correction, valley cold air pool effect, mountain internal boundary layer formation effect, urban land-use effect) in complex terrain areas. In addition, the statistical bias correction of the LDAPS model were efficient in reducing forecast errors of the near-surface temperatures. The new high-resolution downscale model showed better agreement against Korean 584 meteorological monitoring stations than the KMAP, supporting the importance of the new physical and statistical correction methods. The new physical/statistical diagnostic downscale model can be a useful tool in improving near-surface temperature forecasts and diagnostics over complex terrain areas.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.1
• /
• pp.1-8
• /
• 2023

In this paper, we propose a training method of a recommendation learning model that analyzes the battlefield situation and recommends a suitable hypothesis for the current situation. The proposed learning model uses the preference determined by comparing the two hypotheses as a label data to learn which hypothesis best analyzes the current battlefield situation. Our model is based on Siamese neural network architecture which uses the same weights on two different input vectors. The model takes two hypotheses as an input, and learns the priority between two hypotheses while sharing the same weights in the twin network. In addition, a score is given to each hypothesis through the proposed post-processing ranking algorithm, and hypotheses with a high score can be recommended to the commander in charge.

• Journal of the Society of Disaster Information
• /
• v.18 no.4
• /
• pp.786-794
• /
• 2022

In construction sites, effective preventive safety management is required beyond post-processing safety management. Purpose: The purpose of this study is to present a model and analyze the relationship between safety management activities, safety culture key elements, safety, unsafe behavior management, and safety for autonomous and preventive safety management. Method: The relationship was analyzed by applying the survey data to the structural equation, and the path to safety outcomes from exogenous variables was explored and major issues were presented by interpreting the part suggested by the hypothesis verified by the analysis results. Result: As a result of analyzing the preliminary model and the path model, the appropriate model fit was confirmed, and the significant effect of exogenous variables on endogenous variables was confirmed. Conclusion: It is judged that it can contribute to continuously improving safety performance before safety accidents occur through safety management activities, safety and unsafe behavior management, and management of key elements of safety culture.

• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.351-363
• /
• 2023

Visual structural inspections are an inseparable part of post-earthquake damage assessments. With unmanned aerial vehicles (UAVs) establishing a new frontier in visual inspections, there are major computational challenges in processing the collected massive amounts of high-resolution visual data. We propose twin deep learning models that can provide accurate high-resolution structural components and damage segmentation masks efficiently. The traditional approach to cope with high memory computational demands is to either uniformly downsample the raw images at the price of losing fine local details or cropping smaller parts of the images leading to a loss of global contextual information. Therefore, our twin models comprising Trainable Resizing for high-resolution Segmentation Network (TRS-Net) and DmgFormer approaches the global and local semantics from different perspectives. TRS-Net is a compound, high-resolution segmentation architecture equipped with learnable downsampler and upsampler modules to minimize information loss for optimal performance and efficiency. DmgFormer utilizes a transformer backbone and a convolutional decoder head with skip connections on a grid of crops aiming for high precision learning without downsizing. An augmented inference technique is used to boost performance further and reduce the possible loss of context due to grid cropping. Comprehensive experiments have been performed on the 3D physics-based graphics models (PBGMs) synthetic environments in the QuakeCity dataset. The proposed framework is evaluated using several metrics on three segmentation tasks: component type, component damage state, and global damage (crack, rebar, spalling). The models were developed as part of the 2^nd International Competition for Structural Health Monitoring.

• Computers and Concrete
• /
• v.31 no.5
• /
• pp.457-468
• /
• 2023

Buildings, bridges, and dams are examples of civil infrastructure that play an important role in public life. These structures are prone to structural variations over time as a result of external forces that might disrupt the operation of the structures, cause structural integrity issues, and raise safety concerns for the occupants. Therefore, monitoring the state of a structure, also known as structural health monitoring (SHM), is essential. Owing to the emergence of the fourth industrial revolution, next-generation sensors, such as wireless sensors, UAVs, and video cameras, have recently been utilized to improve the quality and efficiency of building forensics. This study presents a method that uses a target-based system to estimate the dynamic displacement and its corresponding dynamic properties of structures using UAV-based video. A laboratory experiment was performed to verify the tracking technique using a shaking table to excite an SDOF specimen and comparing the results between a laser distance sensor, accelerometer, and fixed camera. Then a field test was conducted to validate the proposed framework. One target marker is placed on the specimen, and another marker is attached to the ground, which serves as a stationary reference to account for the undesired UAV movement. The results from the UAV and stationary camera displayed a root mean square (RMS) error of 2.02% for the displacement, and after post-processing the displacement data using an OMA method, the identified natural frequency and damping ratio showed significant accuracy and similarities. The findings illustrate the capabilities and reliabilities of the methodology using UAV to evaluate the dynamic properties of structures.

• 한국정보교육학회:학술대회논문집
• /
• 2021.08a
• /
• pp.77-84
• /
• 2021

The purpose of this study is to propose AI maker coding education as a way to improve computational thinking(CT), which is an essential competence for problem-solving capability in modern society, and to analyze the effectiveness of this education on improving CT in elementary school students. For the research, 5 students from 4th graders and 5 students from 6th graders were recruited, and AI maker coding education was planned in 8 sessions to form classes from basic block coding and maker education to real-life problem solving. To analyze the effectiveness of AI maker coding education, pre- and post-CT examinations were performed. The test results confirmed that AI maker coding education had a significant effect on "abstraction", "algorithm", and "data processing" in the five CT components, and confirmed that there was no correlation in "problem resolution" and "automation". Overall, the average score of all students increased, and the deviation between students decreased, confirming that AI maker coding education was effective in improving CT.

• Journal of Advanced Navigation Technology
• /
• v.27 no.5
• /
• pp.643-648
• /
• 2023

The Korea Augmentation Satellite System is an SBAS system being developed with the goal of providing SoL (Safety of Life) in accordance with ICAO (International Civil Aviation Organization) standards by December 2023. Monitoring the status of the system is essential for the continuous provision of KASS services, and a status monitoring tool should be developed for this purpose. The development of a status monitoring tool was divided into SYSRT (SYStem Real Time monitoring tool), SMSPP (Subsystem Monitoring Statistics tool for Post Processing) depending on the purpose. Tool development was completed through a series of procedures: requirements definition, design, development, and verification. To verify the status monitoring tool, the KASS system's real data (August 2023) were used to verify it, and the results were statistically analyzed to derive operating time and operating rate. It plans to use these tools to support continuous service provision for SoL service starting after this year.