• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.383-392
• /
• 2023

Post-earthquake building condition assessment is crucial for subsequent rescue and remediation and can be automated by emerging computer vision and deep learning technologies. This study is based on an endeavour for the 2nd International Competition of Structural Health Monitoring (IC-SHM 2021). The task package includes five image segmentation objectives - defects (crack/spall/rebar exposure), structural component, and damage state. The structural component and damage state tasks are identified as the priority that can form actionable decisions. A multi-task Convolutional Neural Network (CNN) is proposed to conduct the two major tasks simultaneously. The rest 3 sub-tasks (spall/crack/rebar exposure) were incorporated as auxiliary tasks. By synchronously learning defect information (spall/crack/rebar exposure), the multi-task CNN model outperforms the counterpart single-task models in recognizing structural components and estimating damage states. Particularly, the pixel-level damage state estimation witnesses a mIoU (mean intersection over union) improvement from 0.5855 to 0.6374. For the defect detection tasks, rebar exposure is omitted due to the extremely biased sample distribution. The segmentations of crack and spall are automated by single-task U-Net but with extra efforts to resample the provided data. The segmentation of small objects (spall and crack) benefits from the resampling method, with a substantial IoU increment of nearly 10%.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2020.11a
• /
• pp.1036-1038
• /
• 2020

This paper proposes a novel automated liver segmentation using Multi-Dilated U-Nets. The proposed multidilation segmentation model has the advantage of considering both local and global shapes of the liver image. We use the CT images subject-wise, every 2D image is concatenated to 3D to calculate the IOU score and DICE score. The experimental results on Jeonbuk National University hospital dataset achieves better performance than the conventional U-Net.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.10
• /
• pp.2718-2731
• /
• 2023

This paper presents a method of using syntax and shallow semantic analysis for Vietnamese question generation (QG). Specifically, our proposed technique concentrates on investigating both the syntactic and shallow semantic structure of each sentence. The main goal of our method is to generate questions from a single sentence. These generated questions are known as factoid questions which require short, fact-based answers. In general, syntax-based analysis is one of the most popular approaches within the QG field, but it requires linguistic expert knowledge as well as a deep understanding of syntax rules in the Vietnamese language. It is thus considered a high-cost and inefficient solution due to the requirement of significant human effort to achieve qualified syntax rules. To deal with this problem, we collected the syntax rules in Vietnamese from a Vietnamese language textbook. Moreover, we also used different natural language processing (NLP) techniques to analyze Vietnamese shallow syntax and semantics for the QG task. These techniques include: sentence segmentation, word segmentation, part of speech, chunking, dependency parsing, and named entity recognition. We used human evaluation to assess the credibility of our model, which means we manually generated questions from the corpus, and then compared them with the generated questions. The empirical evidence demonstrates that our proposed technique has significant performance, in which the generated questions are very similar to those which are created by humans.

• Korean Journal of Radiology
• /
• v.22 no.1
• /
• pp.118-130
• /
• 2021

Objective: This study aimed to investigate the blood-brain barrier (BBB) disruption in mild traumatic brain injury (mTBI) patients with post-concussion syndrome (PCS) using dynamic contrast-enhanced (DCE) magnetic resonance (MR) imaging and automatic whole brain segmentation. Materials and Methods: Forty-two consecutive mTBI patients with PCS who had undergone post-traumatic MR imaging, including DCE MR imaging, between October 2016 and April 2018, and 29 controls with DCE MR imaging were included in this retrospective study. After performing three-dimensional T1-based brain segmentation with FreeSurfer software (Laboratory for Computational Neuroimaging), the mean K^trans and v_p from DCE MR imaging (derived using the Patlak model and extended Tofts and Kermode model) were analyzed in the bilateral cerebral/cerebellar cortex, bilateral cerebral/cerebellar white matter (WM), and brainstem. K^trans values of the mTBI patients and controls were calculated using both models to identify the model that better reflected the increased permeability owing to mTBI (tendency toward higher K^trans values in mTBI patients than in controls). The Mann-Whitney U test and Spearman rank correlation test were performed to compare the mean K^trans and v_p between the two groups and correlate K^trans and v_p with neuropsychological tests for mTBI patients. Results: Increased permeability owing to mTBI was observed in the Patlak model but not in the extended Tofts and Kermode model. In the Patlak model, the mean K^trans in the bilateral cerebral cortex was significantly higher in mTBI patients than in controls (p = 0.042). The mean v_p values in the bilateral cerebellar WM and brainstem were significantly lower in mTBI patients than in controls (p = 0.009 and p = 0.011, respectively). The mean K^trans of the bilateral cerebral cortex was significantly higher in patients with atypical performance in the auditory continuous performance test (commission errors) than in average or good performers (p = 0.041). Conclusion: BBB disruption, as reflected by the increased K^trans and decreased v_p values from the Patlak model, was observed throughout the bilateral cerebral cortex, bilateral cerebellar WM, and brainstem in mTBI patients with PCS.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.44 no.5
• /
• pp.55-61
• /
• 2007

In this paper, we propose a three-dimensional(3D) active shape models for medical image segmentation. In order to build a 3D shape model, we need to generate a point distribution model(PDM) and select corresponding landmarks in all the training shapes. The manual determination method, two-dimensional(2D) method, and limited 3D method of landmark correspondences are time-consuming, tedious, and error-prone. In this paper, we generate a 3D statistical shape model using the 3D model generation method of a distance transform and a tetrahedron method for landmarking. After generating the 3D model, we extend the shape model training and gray-level model training of 2D active shape models(ASMs) and we use the integrated modeling process with scale and gray-level models for the appearance profile to represent the local structure. Experimental results are comparable to those of region-based, contour-based methods, and 2D ASMs.

• The Korean Journal of Applied Statistics
• /
• v.18 no.1
• /
• pp.1-13
• /
• 2005

Latent Class model has been considered recently by many researchers and practitioners as a tool for identifying heterogeneous segments or groups in a population, and grouping objects into the segments. In this paper we consider data on prostate cancer patients from Korean National Cancer Institute and propose a method for grouping prostate cancer patients by using latent class Poisson model. A Bayesian approach equipped with a Markov chain Monte Carlo method is used to overcome the limit of classical likelihood approaches. Advantages of the proposed Bayesian method are easy estimation of parameters with their standard errors, segmentation of objects into groups, and provision of uncertainty measures for the segmentation. In addition, we provide a method to determine an appropriate number of segments for the given data so that the method automatically chooses the number of segments and partitions objects into heterogeneous segments.

• The Journal of the Korea Contents Association
• /
• v.13 no.12
• /
• pp.38-47
• /
• 2013

An image coding based on MRC model, a kind of multi-layer image model, first segments a screen image into foreground, mask, and background layers, and then compresses each layer using a codec that is suitable to the layer. The mask layer defines the position of foreground regions such as textual and graphical contents. The colour signal of the foreground (background) region is saved in the foreground (background) layer. The mask layer which contains the segmentation result of foreground and background regions is of importance since its accuracy directly affects the overall coding performance of the codec. This paper proposes a new layer segmentation algorithm for the MRC based image coding. The proposed method extracts text pixels from the background using morphological top hat filtering. The application of white or black top hat transformation to local blocks is controlled by the information of relative brightness of text compared to the background. In the proposed method, the boundary information of text that is extracted from the edge map of the block is used for the robust decision on the relative brightness of text. Simulation results show that the proposed method is superior to the conventional methods.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.4
• /
• pp.51-61
• /
• 2004

In the dental field, the 3D tooth model in which each tooth can be manipulated individually is an essential component for the simulation of orthodontic surgery and treatment. To reconstruct such a tooth model from CT slices, we need to define the accurate boundary of each tooth from CT slices. However, the global threshold method, which is commonly used in most existing 3D reconstruction systems, is not effective for the tooth segmentation in the CT image. In tooth CT slices, some teeth touch with other teeth and some are located inside of alveolar bone whose intensity is similar to that of teeth. In this paper, we propose an image segmentation algorithm based on B-spline curve fitting to produce smooth tooth regions from such CT slices. The proposed algorithm prevents the malfitting problem of the B-spline algorithm by providing accurate initial tooth boundary for the fitting process. This paper proposes an optimal threshold scheme using the intensity and shape information passed by previous slice for the initial boundary generation and an efficient B-spline fitting method based on genetic algorithm. The test result shows that the proposed method detects contour of the individual tooth successfully and can produce a smooth and accurate 3D tooth model for the simulation of orthodontic surgery and treatment.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.43 no.2 s.308
• /
• pp.21-32
• /
• 2006

The license plate recognition (LPR) unit consists of the following core components: plate region segmentation, individual character extraction, and character recognition. Out of the above three components, accuracy in the performance of plate region segmentation determines the overall recognition rate of the LPR unit. This paper proposes an algorithm for segmenting the license plate region on the front or rear of a vehicle in a fast and accurate manner. In the case of the proposed algorithm images are captured on the spot where unmanned monitoring of illegal parking and stowage is performed with a variety of roadway environments taken into account. As a means of enhancing the segmentation performance of the on-the-spot-captured images of license plate regions, the proposed algorithm uses a mathematical model for license plate colors to convert color images into digital data. In addition, this algorithm uses Gaussian smoothing and double threshold to eliminate image noises, one-pass boundary tracing to do region labeling, and MBR to determine license plate region candidates and extract individual characters from the determined license plate region candidates, thereby segmenting the license plate region on the front or rear of a vehicle through a verification process. This study contributed to addressing the inability of conventional techniques to segment the license plate region on the front or rear of a vehicle where the frame of the license plate is damaged, through processing images in a real-time manner, thereby allowing for the practical application of the proposed algorithm.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.2
• /
• pp.33-42
• /
• 2022

In this paper, we proposed a deep learning model based on multi-scale dilated convolution feature fusion for the segmentation of black ice region in road image to send black ice warning to drivers in real time. In the proposed multi-scale dilated convolution feature fusion network, different dilated ratio convolutions are connected in parallel in the encoder blocks, and different dilated ratios are used in different resolution feature maps, and multi-layer feature information are fused together. The multi-scale dilated convolution feature fusion improves the performance by diversifying and expending the receptive field of the network and by preserving detailed space information and enhancing the effectiveness of diated convolutions. The performance of the proposed network model was gradually improved with the increase of the number of dilated convolution branch. The mIoU value of the proposed method is 96.46%, which was higher than the existing networks such as U-Net, FCN, PSPNet, ENet, LinkNet. The parameter was 1,858K, which was 6 times smaller than the existing LinkNet model. From the experimental results of Jetson Nano, the FPS of the proposed method was 3.63, which can realize segmentation of black ice field in real time.