• Journal of the Korea Computer Graphics Society
• /
• v.30 no.1
• /
• pp.11-17
• /
• 2024

Lung cancer segmentation in chest CT images is challenging due to the varying sizes of tumors and the presence of surrounding structures with similar intensity values. To address these issues, we propose a lung cancer segmentation network that incorporates deep supervision and utilizes UNet3+ as the backbone. Additionally, we propose a hybrid lesion focal loss function comprising three components: pixel-based, region-based, and shape-based, which allows us to focus on the smaller tumor regions relative to the background and consider shape information for handling ambiguous boundaries. We validate our proposed method through comparative experiments with UNet and UNet3+ and demonstrate that our proposed method achieves superior performance in terms of Dice Similarity Coefficient (DSC) for tumors of all sizes.

• Journal of IKEEE
• /
• v.27 no.4
• /
• pp.456-463
• /
• 2023

In this paper, we introduce the SSResUNet network model, which integrates the SPADE structure with the U-Net network model for accurate land cover classification using high-resolution satellite imagery without requiring user intervention. The proposed network possesses the advantage of preserving the spatial characteristics inherent in satellite imagery, rendering it a robust classification model even in intricate environments. Experimental results, obtained through training on KOMPSAT-3A satellite images, exhibit superior performance compared to conventional U-Net and U-Net++ models, showcasing an average Intersection over Union (IoU) of 76.10 and a Dice coefficient of 86.22.

• Journal of Forest and Environmental Science
• /
• v.40 no.1
• /
• pp.15-23
• /
• 2024

This research aimed to assess the possibility of detecting forest degradation using time-series satellite imagery and three different deep learning-based change detection techniques. The dataset used for the deep learning models was composed of two sets, one based on surface reflectance (SR) spectral information from satellite imagery, combined with Texture Information (GLCM; Gray-Level Co-occurrence Matrix) and terrain information. The deep learning models employed for land cover change detection included image differencing using the Unet semantic segmentation model, multi-encoder Unet model, and multi-encoder Unet++ model. The study found that there was no significant difference in accuracy between the deep learning models for forest degradation detection. Both training and validation accuracies were approx-imately 89% and 92%, respectively. Among the three deep learning models, the multi-encoder Unet model showed the most efficient analysis time and comparable accuracy. Moreover, models that incorporated both texture and gradient information in addition to spectral information were found to have a higher classification accuracy compared to models that used only spectral information. Overall, the accuracy of forest degradation extraction was outstanding, achieving 98%.

• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.335-349
• /
• 2023

Bridges constantly undergo deterioration and damage, the most common ones being concrete damage and exposed rebar. Periodic inspection of bridges to identify damages can aid in their quick remediation. Likewise, identifying components can provide context for damage assessment and help gauge a bridge's state of interaction with its surroundings. Current inspection techniques rely on manual site visits, which can be time-consuming and costly. More recently, robotic inspection assisted by autonomous data analytics based on Computer Vision (CV) and Artificial Intelligence (AI) has been viewed as a suitable alternative to manual inspection because of its efficiency and accuracy. To aid research in this avenue, this study performs a comparative assessment of different architectures, loss functions, and ensembling strategies for the autonomous segmentation of bridge components and damages. The experiments lead to several interesting discoveries. Nested Reg-UNet architecture is found to outperform five other state-of-the-art architectures in both damage and component segmentation tasks. The architecture is built by combining a Nested UNet style dense configuration with a pretrained RegNet encoder. In terms of the mean Intersection over Union (mIoU) metric, the Nested Reg-UNet architecture provides an improvement of 2.86% on the damage segmentation task and 1.66% on the component segmentation task compared to the state-of-the-art UNet architecture. Furthermore, it is demonstrated that incorporating the Lovasz-Softmax loss function to counter class imbalance can boost performance by 3.44% in the component segmentation task over the most employed alternative, weighted Cross Entropy (wCE). Finally, weighted softmax ensembling is found to be quite effective when used synchronously with the Nested Reg-UNet architecture by providing mIoU improvement of 0.74% in the component segmentation task and 1.14% in the damage segmentation task over a single-architecture baseline. Overall, the best mIoU of 92.50% for the component segmentation task and 84.19% for the damage segmentation task validate the feasibility of these techniques for autonomous bridge component and damage segmentation using RGB images.

• International Journal of Computer Science & Network Security
• /
• v.21 no.9
• /
• pp.203-211
• /
• 2021

Image tampering detection and localization have become an active area of research in the field of digital image forensics in recent times. This is due to the widespread of malicious image tampering. This study presents a new method for image tampering detection and localization that combines the advantages of dilated convolution, residual network, and UNET Architecture. Using the UNET architecture as a backbone, we built the proposed network from two kinds of residual units, one for the encoder path and the other for the decoder path. The residual units help to speed up the training process and facilitate information propagation between the lower layers and the higher layers which are often difficult to train. To capture global image tampering artifacts and reduce the computational burden of the proposed method, we enlarge the receptive field size of the convolutional kernels by adopting dilated convolutions in the residual units used in building the proposed network. In contrast to existing deep learning methods, having a large number of layers, many network parameters, and often difficult to train, the proposed method can achieve excellent performance with a fewer number of parameters and less computational cost. To test the performance of the proposed method, we evaluate its performance in the context of four benchmark image forensics datasets. Experimental results show that the proposed method outperforms existing methods and could be potentially used to enhance image tampering detection and localization.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.37 no.2
• /
• pp.143-149
• /
• 2024

Segmentation of material phases through image analysis is essential for analyzing the microstructure of materials. Micro-CT images exhibit variations in grayscale values depending on the phases constituting the material. Phase segmentation is generally achieved by comparing the grayscale values in the images. In the case of waste concrete used as a recycled aggregate, it is challenging to distinguish between hydrated cement paste and natural aggregates, as these components exhibit similar grayscale values in micro-CT images. In this study, we propose a method for automatically separating the aggregates in concrete, in micro-CT images. Utilizing the Unet-VGG16 deep-learning network, we introduce a technique for segmenting the 2D aggregate images and stacking them to obtain 3D aggregate images. Image filtering is employed to separate aggregate particles from the selected 3D aggregate images. The performance of aggregate segmentation is validated through accuracy, precision, recall, and F1-score assessments.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2013.11a
• /
• pp.638-639
• /
• 2013

Anymon Plus(ver 3.0)은 통합 로그 분석 시스템으로 대용량 로그 및 빅데이터의 실시간 수집 저장 분석할 수 있는 제품(초당 40,000 이벤트 처리)으로서, 방화벽 로그 분석을 통한 비정상 네트워크 행위 탐지, 웹 로그 분석을 통한 사용 패턴 분석, 인터넷 쇼핑몰 사기 주문 분석 및 탐지, 내부 정부 유출 분석 및 탐지 등과 같은 다양한 분야로 응용이 확대되고 있다. 본 논문에서는 보안관련 인프라 로그를 분석하고 예측하여 예상 보안사고 시기에 집중적 경계를 통한 선제적 대응을 모색하기 위해 통계적 이론에 기반한 통합 로그 분석 시스템을 개발하기 위해, 회귀분석 및 시계열 분석이 가능한 예측 엔진 시스템을 설계하고 구현한다.

• Journal of Korean Society on Water Environment
• /
• v.24 no.4
• /
• pp.442-451
• /
• 2008

We analyzed characteristics of rainfall-runoff for the channel of Yongwon area made by a new port construction. And we conducted inundation analysis on the region of lower elevation near the coast. SWMM5 was calibrated with the storm produced by the typhoon Megi from August 19 to August 20 in 2004, and was verified with the storm from August 22 to August 22 in 2004. We performed hydraulic channel routing of Yongwon channel about typhoon Megi from August 19 to August 20 in 2004 by UNET model which is a hydraulic channel routing. The simulated runoff hydrographs were added to the new stream as lateral inflow hydrographs and a watershed runoff hydrograph was the upstream boundary condition. The downstream boundary condition data were estimated by the measured stage hydrographs. The maximum stage that was calculated by hydraulic channel routing was higher than the levee of inundated region in typhoon Megi. Thus we can suppose an inundation to have been occurred. We performed inundation analysis about typhoon Megi from August 19 to August 20 in 2004 and flood discharge of return period 10~150 years. And we estimated each inundation area. The inundation areas by return periods of storms were estimated by 3.4~5.7 ha. The causes of inundation are low heights of levee crests (D.L. 2.033~2.583 m), storm surges induced by typhoons and reverse flow through the coastal sewers (D.L. -0.217~0.783 m). A result of this study can apply to establish countermeasure of a flood disaster in Yongwon.

• Journal of Broadcast Engineering
• /
• v.28 no.3
• /
• pp.267-274
• /
• 2023

There is a problem that the size of the dataset is insufficient due to the limitation of the collection of the medical image public data, so there is a possibility that the existing studies are overfitted to the public dataset. In this paper, we compare the performance of eight (Unet, X-Net, HarDNet, SegNet, PSPNet, SwinUnet, 3D-ResU-Net, UNETR) medical image semantic segmentation models to revalidate the superiority of existing models. Anatomical Tracings of Lesions After Stroke (ATLAS) V1.2, a public dataset for stroke diagnosis, is used to compare the performance of the models and the performance of the models in ATLAS V2.0. Experimental results show that most models have similar performance in V1.2 and V2.0, but X-net and 3D-ResU-Net have higher performance in V1.2 datasets. These results can be interpreted that the models may be overfitted to V1.2.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.31 no.6B
• /
• pp.523-529
• /
• 2011

A tidal river is a river affected by tides causing the water level to rise and fall. In the river, the velocity induced by tides at a dry season can be a more dominant factor than that at a flood season in designing hydraulic structures. In this study, flow velocities and water levels were observed in the downstream of Han River at the dry season, and they are used for verification of numerically predicted results. The behaviors of flow were also analyzed by using UNET, a commercial model managing unsteady flows. To estimate the roughness coefficients of the tidal river reach near the Shingok submerged weir, a statistical method is employed. In the method, the discrepancy ratio between the measured and calculated water levels was implemented.