• 정보처리학회논문지D
• /
• 제12D권4호
• /
• pp.637-646
• /
• 2005

키워드를 통한 웹 검색 결과의 분류와 같은 후처리가 요구되는 문서 분류 문제에서, 기존의 문서 분류 또는 클러스터링 알고리즘을 적용하는 데에는 많은 문제가 있다 그 중에서 고려해야 할 가장 심각한 두 가지 문제가 있다. 첫째는 전문가가 관여하여 범주를 선정하는 문제이고, 둘째는 문서분류에 소요되는 수행시간이 긴 문제이다. 따라서 본 논문에서는 이행적 폐쇄 트리를 이용하여 문서 유사도 계산 횟수를 크게 줄이고, 정확도의 희생을 최소화하면서 신속한 처리가 가능한 새로운 웹 문서 클러스터링 기법을 제안하다. 또한, 제안된 기법의 효율성을 검증하기 위하여 기존의 알고리즘과 비교 평가 및 분석한다.

• ETRI Journal
• /
• 제32권3호
• /
• pp.380-389
• /
• 2010

This paper presents a new edge-protection algorithm and its very large scale integration (VLSI) architecture for block artifact reduction. Unlike previous approaches using block classification, our algorithm utilizes pixel classification to categorize each pixel into one of two classes, namely smooth region and edge region, which are described by the edge-protection maps. Based on these maps, a two-step adaptive filter which includes offset filtering and edge-preserving filtering is used to remove block artifacts. A pipelined VLSI architecture of the proposed deblocking algorithm for HD video processing is also presented in this paper. A memory-reduced architecture for a block buffer is used to optimize memory usage. The architecture of the proposed deblocking filter is verified on FPGA Cyclone II and implemented using the ANAM 0.25 ${\mu}m$ CMOS cell library. Our experimental results show that our proposed algorithm effectively reduces block artifacts while preserving the details. The PSNR performance of our algorithm using pixel classification is better than that of previous algorithms using block classification.

• 한국군사과학기술학회지
• /
• 제21권2호
• /
• pp.158-165
• /
• 2018

Recently there is a great interest in the bi-static sonar. However, since the transmitter and the receiver operate on different platforms, it may be necessary to operate the system in a non-cooperative mode. In this situation, the detection and localization performance are limited. Therefore, it is necessary to classify the received pulse from the transmitter to overcome the performance limitation. In this paper, we proposed a robust automatic pulse classification method that can be applied to real systems. The proposed method eliminates the effects of noise and multipath propagation through post-processing and improves the pulse classification performance. We also verified the proposed method through the sea experimental data.

• 한국음향학회지
• /
• 제30권8호
• /
• pp.461-467
• /
• 2011

제한된 대역폭을 효율적으로 사용하기 위해서 도입된 가변 전송률은 먼저 신호의 정확한 분류를 필요로 한다. 특히 멀티미디어 서비스가 보편화 되면서 음성/음악 신호 분류의 중요성도 높아지게 되었다. 음성/음악 분류기 중, 서포트벡터머신 (SVM)을 이용한 분류기는 높은 분류 정확도로 주목받고 있다. 그러나 SVM는 많은 계산량과 저장 공간을 요구하므로 효율적인 구현이 요구되며, 특히 임베디드 시스템과 같이 자원이 제한 적인 경우에는 더욱 그러하다. 본 논문에서는 먼저 SVM을 이용한 음성/음악 분류기의 임베디드 시스템으로의 구현을 실행시간과 에너지소비의 관점에서 분석하고, 효율적인 구현을 위한 두가지 방법들을 제안한다. 서포트벡터의 판별결과에의 기여도를 바탕으로 기여도가 낮은 벡터들을 제외하는 방법과, 음성/음악 신호에 기본적으로 존재하는 각 프레임간의 상관관계를 이용하여 입력신호의 일부를 건너뛰는 방법이다. 이 기법들은 SVM의 학습 시 사용되는 다른 최적화 기법에 관계없이 적용이 가능하며, 실험을 통해 분류의 정확도, 실행시간, 그리고 에너지소비의 관점에서 그 성능을 증명하였다.

• Journal of Gastric Cancer
• /
• 제11권3호
• /
• pp.173-179
• /
• 2011

Purpose: It is difficult to obtain biopsies from gastrointestinal stromal tumors (GISTs) prior to surgery because GISTs are submucoal tumors, despite being the most common nonepithelial neoplasms of the gastrointestinal tract. Unlike anatomic imaging techniques, PET-CT, which is a molecular imaging tool, can be a useful technique for assessing tumor activity and predicting the malignant potential of certain tumors. Thus, we aimed to evaluate the usefulness of PET-CT as a pre-operative prognostic factor for GISTs by analyzing the correlation between the existing post-operative prognostic factors and the maximum SUV uptake (SUVmax) of pre-operative 18F-fluoro-2-deoxyglucose (FDG) PET-CT. Materials and Methods: The study was conducted on 26 patients who were diagnosed with gastric GISTs and underwent surgery after being examined with pre-operative FDG PET-CT. An analysis of the correlation bewteen (i) NIH risk classification and the Ki-67 proliferation index, which are post-operative prognostic factors, and (ii) the SUVmax of PET-CT, which is a pre-operative prognostic factor, was performed. Results: There were significant correlations between (i) SUVmax and (ii) Ki-67 index, tumor size, mitotic count, and NIH risk group (r=0.854, 0.888, 0.791, and 0.756, respectively). The optimal cut-off value for SUVmax was 3.94 between "low-risk malignancy" and "high-risk malignancy" groups. The sensitivity and specificity of SUVmax for predicting the risk of malignancy were 85.7% and 94.7%, respectively. Conclusions: The SUVmax of PET-CT is associated with Ki-67 index, tumor size, mitotic count, and NIH classification. Therefore, it is believed that PET-CT is a relatively safe, non-invasive diagnostic tool for assessing malignant potential pre-operatively.

• IEIE Transactions on Smart Processing and Computing
• /
• 제4권4호
• /
• pp.195-201
• /
• 2015

For median filtering (MF) detection in altered digital images, this paper presents a new feature vector that is formed from autoregressive (AR) coefficients via an AR model of the gradients between the neighboring row and column lines in an image. Subsequently, the defined 10-D feature vector is trained in a support vector machine (SVM) for MF detection among forged images. The MF classification is compared to the median filter residual (MFR) scheme that had the same 10-D feature vector. In the experiment, three kinds of test items are area under receiver operating characteristic (ROC) curve (AUC), classification ratio, and minimal average decision error. The performance is excellent for unaltered (ORI) or once-altered images, such as $3{\times}3$ average filtering (AVE3), QF=90 JPEG (JPG90), 90% down, and 110% up to scale (DN0.9 and Up1.1) images, versus $3{\times}3$ and $5{\times}5$ median filtering (MF3 and MF5, respectively) and MF3 and MF5 composite images (MF35). When the forged image was post-altered with AVE3, DN0.9, UP1.1 and JPG70 after MF3, MF5 and MF35, the performance of the proposed scheme is lower than the MFR scheme. In particular, the feature vector in this paper has a superior classification ratio compared to AVE3. However, in the measured performances with unaltered, once-altered and post-altered images versus MF3, MF5 and MF35, the resultant AUC by 'sensitivity' (TP: true positive rate) and '1-specificity' (FN: false negative rate) is achieved closer to 1. Thus, it is confirmed that the grade evaluation of the proposed scheme can be rated as 'Excellent (A)'.

• 대한조선학회논문집
• /
• 제58권5호
• /
• pp.303-313
• /
• 2021

Recently, quality control of the Liquefied Natural Gas Carrier (LNGC) cargo hold and block-erection interference areas using 3D scanners have been performed, focusing on large shipyards and the international association of classification societies. In this study, as a part of the research on LNGC cargo hold quality management advancement, a study on deep-learning-based scaffolding system 3D point cloud object detection and post-processing were conducted using a LNGC cargo hold 3D point cloud. The scaffolding system point cloud object detection is based on the PointNet deep learning architecture that detects objects using point clouds, achieving 70% prediction accuracy. In addition, the possibility of improving the accuracy of object detection through parameter adjustment is confirmed, and the standard of Intersection over Union (IoU), an index for determining whether the object is the same, is achieved. To avoid the manual post-processing work, the object detection architecture allows automatic task performance and can achieve stable prediction accuracy through supplementation and improvement of learning data. In the future, an improved study will be conducted on not only the flat surface of the LNGC cargo hold but also complex systems such as curved surfaces, and the results are expected to be applicable in process progress automation rate monitoring and ship quality control.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제12권8호
• /
• pp.333-340
• /
• 2023

기업 부도의 효율적인 예측은 금융기관의 적절한 대출 결정과 여신 부실률 감소 측면에서 중요한 부분이다. 많은 연구에서 인공지능 기술을 활용한 분류모델 연구를 진행하였다. 금융 산업 특성상 새로운 예측 모델의 성능이 우수하더라도 어떤 근거로 결과를 출력했는지 직관적인 설명이 수반되어야 한다. 최근 미국, EU, 한국 등 에서는 공통적으로 알고리즘의 설명요구권을 제시하고 있어 금융권 AI 활용에 투명성을 확보하여야 한다. 본 논문에서는 외부에 오픈된 기업부도 데이터를 활용하여 인공지능 기반의 해석 가능한 분류 예측 모델을 제안하였다. 먼저 데이터 전처리 작업, 5겹 교차검증 등을 수행하고 로지스틱 회귀, SVM, XGBoost, LightGBM 등 10가지 지도학습 분류모델 최적화를 통해 분류 성능을 비교하였다. 그 결과 LightGBM이 가장 우수한 모델로 확인되었고, 설명 가능한 인공지능 기법인 SHAP을 적용하여 부도예측 과정에 대한 사후 설명을 제공하였다.

• 한국임상약학회지
• /
• 제28권1호
• /
• pp.57-64
• /
• 2018

Background: Patients experience significant differences in aspects of mortality, quality of life, and costs between during the year of receiving liver transplant (LT) and the subsequent years (post-LT). This study aimed to estimate the medical utilization and cost of LT for patients compared to post-LT patients by using a recent National Patient Sample (NPS) data provided by the Korean Health Insurance Review and Assessment Service (HIRA). Methods: This study used a subset of the 2015 HIRA-NPS. Patient claims data that included Z944 (Korean Standard Classification of Diseases code for LT status) were selected. Within the selected data, LT patients were identified based on whether the national health insurance number code of Q80 (procedure code for LT surgery) was included, and they were compared to post-LT patients. Results: In the analysis, 330 patients were included. The average cost per patient was $90,066{\pm}36,959$ thousand KRW and $10,557{\pm}9,668$ thousand KRW for LT and post-LT patients, respectively. Especially, LT patients' costs for injection/procedure, surgery/treatment, and examination were higher than other costs, being $35,983{\pm}18,115$ thousand KRW, $28,246{\pm}9,408$ thousand KRW, and $12,131{\pm}6,604$ thousand KRW, respectively. For inpatients, the average number of hospitalized days was $63.5{\pm}66.0$ days for LT patients and $22.3{\pm}35.1$ days for post-LT patients. Conclusion: Compared to post-LT patients, LT patients had higher costs, especially for injection/procedure, surgery/treatment, and examination. Additionally, the LT group had longer hospitalization duration and higher costs for their hospital admission, whereas they did not show a significant difference in number of visits and medical costs for outpatient-care.

• 대한족부족관절학회지
• /
• 제17권1호
• /
• pp.28-33
• /
• 2013

Purpose: In previous study, the frequency of osteoarthritis and discomfort were high in ankle fracture-dislocation but detail results about ankle fracture-dislocation has rarely been investigated. In this study, we retrospectively analyze the outcome of the operative treatment of ankle fracture with dislocation for over-2 years follow up. Materials and Methods: There were 47 cases of ankle fracture-dislocation in our hospital from March 2007 to May 2010. We investigated 20 patients who underwent operation and were possible for over-2 years follow up. The result was estimated with the direction of dislocation, fracture type, the time of bone union and post-traumatic osteoarthritis with plain radiologic images. In clinical assessment, we statistically evaluated the function and pain through AOFAS score and Olerud & Molander scoring system. Results: By Lauge-Hansen classification, there were 13 cases(65%) of pronation-external rotation and 6 cases(30%) of supination-external rotation, 1 case(5%) of supination-adduction. AOFAS score was 85.5, and Olerud & Molander score was "excellent" in 8 cases, "good" in 5 cases, "fair" in 3 cases and "poor" in 4 cases. Postoperative complications in 4 cases revealed post-traumatic arthritis. All kinds of lateral dislocation of ankle fracture was 15 cases and the most common. Of these, all 3 cases, anterolateral dislocation showed post-traumatic osteoarthritis. Conclusion: In ankle fracture-dislocation, post-traumatic osteoarthritis occurred in 4 of 20 patient(20%). Especially, the possibility of post-traumatic osteoarthritis was more in cases of anterolateral or lateral dislocation. So, it must be needed that deliberate examination, for example, preoperative MRI and sufficient explanation to patient. Also, we have to follow up the patients carefully.