Search

Your search keyword '"Park, Seunghee"' showing total 413 results
Start Over Author "Park, Seunghee"
413 results on '"Park, Seunghee"'

1. Low-Light Image Enhancement Framework for Improved Object Detection in Fisheye Lens Datasets

Author

Tran, Dai Quoc, Aboah, Armstrong, Jeon, Yuntae, Shoman, Maged, Park, Minsoo, and Park, Seunghee

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

This study addresses the evolving challenges in urban traffic monitoring detection systems based on fisheye lens cameras by proposing a framework that improves the efficacy and accuracy of these systems. In the context of urban infrastructure and transportation management, advanced traffic monitoring systems have become critical for managing the complexities of urbanization and increasing vehicle density. Traditional monitoring methods, which rely on static cameras with narrow fields of view, are ineffective in dynamic urban environments, necessitating the installation of multiple cameras, which raises costs. Fisheye lenses, which were recently introduced, provide wide and omnidirectional coverage in a single frame, making them a transformative solution. However, issues such as distorted views and blurriness arise, preventing accurate object detection on these images. Motivated by these challenges, this study proposes a novel approach that combines a ransformer-based image enhancement framework and ensemble learning technique to address these challenges and improve traffic monitoring accuracy, making significant contributions to the future of intelligent traffic management systems. Our proposed methodological framework won 5th place in the 2024 AI City Challenge, Track 4, with an F1 score of 0.5965 on experimental validation data. The experimental results demonstrate the effectiveness, efficiency, and robustness of the proposed system. Our code is publicly available at https://github.com/daitranskku/AIC2024-TRACK4-TEAM15.

Published
2024

2. Chapter Deep Learning Based Pose Estimation of Scaffold Fall Accident Safety Monitoring

Author

Lee, Seungsoo, Son, Seongwoo, Aung, Pa Pa Win, Park, Minsoo, and Park, Seunghee

Subjects
deep learning, pose estimation, keypoint angle calculate, construction site safe monitoring, falls from heights, thema EDItEUR::U Computing and Information Technology
Abstract

According to the Ministry of Manpower, falling and slipping accidents are one of the most common accidents in addition, falls from heights (FFH), including accidents during scaffolding work, are still a major cause of death in the construction industry. Regular safety checks are currently being carried out on construction sites, but scaffold-related accidents continue to occur. Sensing technology is being attempted in many industrial sites for safety monitoring, but there are still limitations in terms of the cost of sensors and object detection, which are limited to certain risks. Therefore, this paper proposes a deep learning-based pose estimation approach to identify the risk of falling during scaffolding work in the construction industry. Through analysis of the correlation between unstable behavior during scaffold work and the angle of keypoints of workers, the proposed approach demonstrates the ability to detect the risk of falling. The proposed approach can prevent falling accidents not only by detecting construction site workers, but also by detecting specific risky behaviors. In addition, in limited work environments other than scaffolding work, the information on unstable behavior can be provided to safety managers who may not be aware of the risk, thus contributing to preventing falling accidents

Published
2023
Full Text
View/download PDF

3. Chapter Deep Learning-Based Pose Estimation for Identifying Potential Fall Hazards of Construction Worker

Author

Lee, Seungsoo, Choi, Woonggyu, Park, Minsoo, Jeon, Yuntae, Quoc Tran, Dai, and Park, Seunghee

Subjects
deep learning, keypoint detection, pose estimation, computer vision, construction site safe, thema EDItEUR::U Computing and Information Technology::UT Computer networking and communications::UTV Virtualization
Abstract

Fall from height (FFH) is one of the major causes of injury and fatalities in construction industry. Deep learning-based computer vision for safety monitoring has gained attention due to its relatively lower initial cost compared to traditional sensing technologies. However, a single detection model that has been used in many related studies cannot consider various contexts at the construction site. In this paper, we propose a deep learning-based pose estimation approach for identifying potential fall hazards of construction workers. This approach can relatively increase the accuracy of estimating the distance between the worker and the fall hazard area compared to the existing methods from the experimental results. Our proposed approach can improve the robustness of worker location estimation compared to existing methods in complex construction site environments with obstacles that can obstruct the worker's position. Also, it is possible to provide information on whether a worker is aware of a potential fall risk area. Our approach can contribute to preventing FFH by providing access information to fall risk areas such as construction site openings and inducing workers to recognize the risk area even in Inattentional blindness (IB) situations

Published
2023
Full Text
View/download PDF

4. Chapter Identifying Hazards in Construction Sites Using Deep Learning-Based Multimodal with CCTV Data

Author

Son, Seongwoo, Quoc Tran, Dai, Jeon, Yuntae, Park, Minsoo, and Park, Seunghee

Subjects
deep learning, multimodal, multiCCTV, synthetic data, pointcloud, thema EDItEUR::U Computing and Information Technology
Abstract

The use of closed-circuit television (CCTV) for safety monitoring is crucial for reducing accidents in construction sites. However, the majority of currently proposed approaches utilize single detection models without considering the context of CCTV video inputs. In this study, a multimodal detection, and depth map estimation algorithm utilizing deep learning is proposed. In addition, the point cloud of the test site is acquired using a terrestrial laser scanning scanner, and the detected object's coordinates are projected into global coordinates using a homography matrix. Consequently, the effectiveness of the proposed monitoring system is enhanced by the visualization of the entire monitored scene. In addition, to validate our proposed method, a synthetic dataset of construction site accidents is simulated with Twinmotion. These scenarios are then evaluated with the proposed method to determine its precision and speed of inference. Lastly, the actual construction site, equipped with multiple CCTV cameras, is utilized for system deployment and visualization. As a result, the proposed method demonstrated its robustness in detecting potential hazards on a construction site, as well as its real-time detection speed

Published
2023
Full Text
View/download PDF

6. Skip-Connected Neural Networks with Layout Graphs for Floor Plan Auto-Generation

Author

Jeon, Yuntae, Tran, Dai Quoc, and Park, Seunghee

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

With the advent of AI and computer vision techniques, the quest for automated and efficient floor plan designs has gained momentum. This paper presents a novel approach using skip-connected neural networks integrated with layout graphs. The skip-connected layers capture multi-scale floor plan information, and the encoder-decoder networks with GNN facilitate pixel-level probability-based generation. Validated on the MSD dataset, our approach achieved a 93.9 mIoU score in the 1st CVAAD workshop challenge. Code and pre-trained models are publicly available at https://github.com/yuntaeJ/SkipNet-FloorPlanGe.

Published
2023

37. TLS-Point Clouding-3D Shape Deflection Monitoring

Author

Cha, Gichun, Yu, Byungjoon, Park, Sehwan, Park, Seunghee, Kacprzyk, Janusz, Series Editor, Pal, Nikhil R., Advisory Editor, Bello Perez, Rafael, Advisory Editor, Corchado, Emilio S., Advisory Editor, Hagras, Hani, Advisory Editor, Kóczy, László T., Advisory Editor, Kreinovich, Vladik, Advisory Editor, Lin, Chin-Teng, Advisory Editor, Lu, Jie, Advisory Editor, Melin, Patricia, Advisory Editor, Nedjah, Nadia, Advisory Editor, Nguyen, Ngoc Thanh, Advisory Editor, Wang, Jun, Advisory Editor, Arai, Kohei, editor, and Kapoor, Supriya, editor

Published
2020
Full Text
View/download PDF

39. Leveraging Semisupervised Learning for Domain Adaptation: Enhancing Safety at Construction Sites through Long-Tailed Object Detection.

Author

Tran, Dai Quoc, Jeon, Yuntae, Aboah, Armstrong, Bak, Jinyeong, Park, Minsoo, and Park, Seunghee

Subjects
BUILDING sites, COMPUTER vision, COMPUTER performance, CLOSED-circuit television, CONSTRUCTION workers, SUPERVISED learning, DEEP learning
Abstract

The advancement of deep learning has led to a growing demand and increase in research on computer vision–based construction site monitoring for improved safety and operational efficiency. These methods largely depend on supervised learning, requiring labeled data for optimal performance. However, when applied to new construction sites with varied environmental conditions, the effectiveness of these models is often compromised. Additionally, highly imbalanced object class distributions in the data sets, known as long-tailed objects, presents significant challenges during model training, considerably impacting performance. Recognizing this crucial gap in the field, this study proposes a novel approach to improve safety and operational efficiency at construction sites by leveraging a semisupervised learning approach for domain adaptation in long-tailed object detection. The method addresses the challenges of unbalanced class distribution and environmental variability in construction site monitoring, which often degrade the performance of computer vision models. By employing semisupervised learning, both labeled and unlabeled data are utilized in domain adaptation to unseen construction sites, considering both image-and object-level noise, thereby enhancing the model's adaptability to diverse working conditions. Based on the detection results, a risk scenario detection algorithm is also introduced for construction vehicles and workers. The efficacy of the proposed approach was validated through extensive experiments conducted on a comprehensive data set sourced from AIHub and CrowdHuman, in addition to actual self-labeled closed-circuit television (CCTV) data comprising 500 videos from construction sites' CCTV cameras. The evaluations revealed that the proposed method significantly outperforms conventional semisupervised learning by 9.76% on mean average precision for construction vehicle detection and by 3% for the worker detection model, paving the way for advanced construction site monitoring systems that ensure a safer and more efficient working environment. [ABSTRACT FROM AUTHOR]

Published
2025
Full Text
View/download PDF

43. Magnetic Hall Sensor-Generated MFL-Based Improved Metal Loss Detection and SVM-Based Quantification for Ferromagnetic Structural Components

Author

Emagnenehe Yigzew, Fitsum, Kim, Hansun, Asefa Beyene, Daniel, and Park, Seunghee

Abstract

This study focuses on the assessment of magnetic Hall sensors’ orientation-based application through magnetic flux leakage excitation data from a sample target ferromagnetic structure. Initially, we designed a prototype of a rectangular sensor head with ten channel compartments, each equipped with Hall magnetic sensors, to conduct contact-based experiments on a ferromagnetic structural component. The target specimen, a steel plate, was subjected to a strong magnetic field using a Neodymium-35 magnet. Various-sized metal loss anomalies were deliberately induced in the target plate structure. Upon scanning the target specimen at a constant speed, the Hall magnetic sensor collected and stored data on both intact and damaged magnetic flux. To enhance accuracy and extract essential features, the raw flux leakage input underwent sequential signal adjustment steps. Signal averaging was employed to improve the overall data gathered from repeated experiments. The magnetic induction signal curve was analyzed to observe the nature of flux leakage in the magnetic Hall sensors. Frequency analysis and sorted amplitude analysis were conducted to gain a deeper understanding of significant features. Following box plot-based outlier identification, a low-pass filter and fast Fourier transform-based filtering were applied to eliminate unwanted flux leakage. Data detrending and signal-to-noise ratio analysis were employed to assess the baseline variation filtering results for improving the original flux leakage data. To allocate defects, a three-step iteration-based wavelet decomposition was proposed, successfully determining the defect region. Anomaly size estimation was carried out using a support vector machine, yielding promising results.

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results