• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권9호
• /
• pp.4412-4428
• /
• 2018

The purpose of this work is to solve the problem of representing an entire video using Convolutional Neural Network (CNN) features for human action recognition. Recently, due to insufficient GPU memory, it has been difficult to take the whole video as the input of the CNN for end-to-end learning. A typical method is to use sampled video frames as inputs and corresponding labels as supervision. One major issue of this popular approach is that the local samples may not contain the information indicated by the global labels and sufficient motion information. To address this issue, we propose a binary hashing method to enhance the local feature extractors. First, we extract the local features and aggregate them into global features using maximum/minimum pooling. Second, we use the binary hashing method to capture the motion features. Finally, we concatenate the hashing features with global features using different normalization methods to train the classifier. Experimental results on the JHMDB and MPII-Cooking datasets show that, for these new local features, binary hashing mapping on the sparsely sampled features led to significant performance improvements.

• 한국멀티미디어학회논문지
• /
• 제11권12호
• /
• pp.1635-1648
• /
• 2008

The extraction of important features in cancer cell image analysis is a key process in grading renal cell carcinoma. In this study, we applied three-dimensional (3D) texture feature extraction methods to cell nuclei images and evaluated the validity of them for computerized cell nuclei grading. Individual images of 2,423 cell nuclei were extracted from 80 renal cell carcinomas (RCCs) using confocal laser scanning microscopy (CLSM). First, we applied the 3D texture mapping method to render the volume of entire tissue sections. Then, we determined the chromatin texture quantitatively by calculating 3D gray-level co-occurrence matrices (3D GLCM) and 3D run length matrices (3D GLRLM). Finally, to demonstrate the suitability of 3D texture features for grading, we performed a discriminant analysis. In addition, we conducted a principal component analysis to obtain optimized texture features. Automatic grading of cell nuclei using 3D texture features had an accuracy of 78.30%. Combining 3D textural and 3D morphological features improved the accuracy to 82.19%. As a comparative study, we also performed a stepwise feature selection. Using the 4 optimized features, we could obtain more improved accuracy of 84.32%. Three dimensional texture features have potential for use as fundamental elements in developing a new nuclear grading system with accurate diagnosis and predicting prognosis.

• Design & Manufacturing
• /
• 제16권3호
• /
• pp.50-57
• /
• 2022

The qualities of the products produced by injection molding are strongly influenced by the process variables of the injection molding machine set by the engineer. It is very difficult to predict the qualities of the injection molded product considering the stochastic nature of the manufacturing process, since the processing conditions have a complex impact on the quality of the injection molded product. It is recognized that the artificial neural network(ANN) is capable of mapping the intricate relationship between the input and output variables very accurately, therefore, many studies are being conducted to predict the relationship between the results of the product and the process variables using ANN. However in the condition of a small number of data sets, the predicting performance and robustness of the ANN model could be reduced due to too many input variables. In the present study, the ANN model that predicts the length of the injection molded product for multiple combinations of process variables was developed. And the accuracy of each ANN model was compared for 8 process variables and 4 important process inputs that were determined by the feature selection. Based on the comparison, it was verified that the performance of the ANN model increased when only 4 important variables were applied.

• 대한원격탐사학회지
• /
• 제24권2호
• /
• pp.117-124
• /
• 2008

지형지물은 각각의 특징적 요인을 내포하고 있다. 이 특징적 요인들은, 공간해상도에 따라 정도의 차이가 있겠지만, 수집된 위성영상에도 반영된다. 이러한 요인들 중에서는 영상분류에 활용될 경우 영상 분류의 정확도를 높혀주고, 때로는 이것이 거의 물체인식의 수준까지 기여할 수 있는 것들이 있다. 이 연구에서는 텍스춰 및 지형지물의 배열에 있어서 특징적 현상을 보이는 비닐하우스를 대상으로 spatial auto-corelation 개념을 기반으로 자동적으로 이를 인지하는 방법을 개발하였다. 사용된 알고리즘은 디지타이징과 같은 사람의 직접적인 개입이 없이 자동화된 방법으로 비닐하우스의 특정한 패턴이 반복적으로 나타나는 것을 감지할 수 있도록 개발되었다. 패틴의 인식에 더하여 비닐하우스의 기하학적 모양을 고려하는 방법도 도입하였다. 그럼으로써 비닐하우스의 추출에 단순히 화소 단위의 분석이 아닌 보다 객체지향적인 방법으로 비닐하우스를 추출하도록 하였다. 개발된 방법을 제주지역의 IKONOS에 적용시켜 본 결과 연구대상지역내의 비닐하우스가 매우 정확하게 적출되었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권10호
• /
• pp.2768-2787
• /
• 2023

As an important research direction of the application of computer science in the medical field, the automatic generation technology of radiology report has attracted wide attention in the academic community. Because the proportion of normal regions in radiology images is much larger than that of abnormal regions, words describing diseases are often masked by other words, resulting in significant feature loss during the calculation process, which affects the quality of generated reports. In addition, the huge difference between visual features and semantic features causes traditional multi-modal fusion method to fail to generate long narrative structures consisting of multiple sentences, which are required for medical reports. To address these challenges, we propose a multi-feature optimization Transformer (MFOT) for generating radiology reports. In detail, a multi-dimensional mapping attention (MDMA) module is designed to encode the visual grid features from different dimensions to reduce the loss of primary features in the encoding process; a feature pre-fusion (FP) module is constructed to enhance the interaction ability between multi-modal features, so as to generate a reasonably structured radiology report; a detail enhanced attention (DEA) module is proposed to enhance the extraction and utilization of key features and reduce the loss of key features. In conclusion, we evaluate the performance of our proposed model against prevailing mainstream models by utilizing widely-recognized radiology report datasets, namely IU X-Ray and MIMIC-CXR. The experimental outcomes demonstrate that our model achieves SOTA performance on both datasets, compared with the base model, the average improvement of six key indicators is 19.9% and 18.0% respectively. These findings substantiate the efficacy of our model in the domain of automated radiology report generation.

• 한국측량학회지
• /
• 제10권2호
• /
• pp.63-71
• /
• 1992

디지탈 지도 데이타베이스는 다중 축적의 개념을 포함하여 여러 가지 목적을 두고 제작되며 단일 축적으로만 사용하기 위해 Base Map을 구축하는 사례는 극히 보기 드믄 현상이라고 할 수 있다. 따라서 지도의 일반화와 다중표현에 대한 Line의 단순화 처리에 있어서 가장 중요한 문제는 일반화된 그래픽 데이타의 정확도와 인식도를 모두 부여하기 위해 Base Map 상의 정보를 단순화하기 위해 설정하는 허용범위를 디지틸 화일내에서 Feature의 형태에 따라 수정이 가능하도록 하는 것이다. 본 연구에서는 하나의 디지털 화일내에서 다양한 축척상으로 수행되는 Line의 단순화에 대한 여러가지 알고리즘을 고찰하였으며, 지도의 표현상에 변화를 줄 수 있는 선형성 Feature 별로 축척에 따른 규칙을 설정하였다. 수치화된 line 데이타 사이의 상관성을 분석하기 위하여 2가지 변형량을 측정하여 5가지 알고리즘에 대한 평가를 시도하였다. 데이타의 분석결과 Douglas-Peucker 알고리즘이 단순화 후의 변형량에 있어 가장 작은 영향을 받음을 알 수 있었다. 이러한 연구 결과로부터 디지탈 화일을 소축척으로 표현하기 위해 단순화를 실시할 경우 내부적으로 지니고 있어야 하는 기하학적인 항목으로서 그 크기와 변동량에 대한 수치적인 안을 제시함에 따라 지도의 단순화에 대한 가능성을 입증하였다.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2004년도 Proceedings of ISRS 2004
• /
• pp.344-346
• /
• 2004

IKONOS 1m satellite imagery is particularly well suited for 3-D feature extraction and 1 :5,000 scale topographic mapping. Because the image line and sample calculated by given RPCs have the error of more than 11m, in order to be able to perform feature extraction and topographic mapping, rational polynomial coefficients(RPCs) camera model that are derived from the very complex IKONOS sensor model to describe the object-image geometry must be refined by several Ground Control Points(GCPs). This paper presents a quantitative evaluation of the geometric accuracy that can be achieved with IKONOS imagery by refining the offset and scaling factors of RPCs using several GCPs. If only two GCPs are available, the offsets and scale factors of image line and sample are updated. If we have more than three GCPs, four parameters of the offsets and scale factors of image line and sample are refined first, and then six parameters of the offsets and scale factors of latitude, longitude and height are updated. The stereo images acquired by IKONOS satellite are tested using six ground points. First, the RPCs model was refined using 2 GCPs and 4 check points acquired by GPS. The results from IKONOS stereo images are reported and these show that the RMSE of check point acquired from left images and right are 1.021m and 1.447m. And then we update the RPCs model using 4 GCPs and 2 check points. The RMSE of geometric accuracy is 0.621 m in left image and 0.816m in right image.

• 전자공학회논문지CI
• /
• 제47권4호
• /
• pp.25-33
• /
• 2010

본 논문에서는 단일 카메라를 이용하여 얼굴의 움직임 정보를 추정하고 3차원 모델을 합성하기 위한 기법을 제안한다. 먼저 단일 카메라 입력 영상에서 사용자의 얼굴 영역 특징 점 취득을 위한 4개의 하부 이미지를 획득한다. 획득된 4개의 하부 이미지를 템플릿으로 사용하여 사용자 얼굴 영역의 정보를 추출하며, 이들 4개의 특징 점을 사용하여 사용자 얼굴과 카메라 영상 평면 사이의 사영 관계를 계산한다. 취득된 카메라 행렬로부터 얼굴의 움직임 정보인 이동과 회전 성분을 추정할 수 있으며, 이를 기반으로 3차원 모델의 자세 정보를 설정한 다음 이를 사용자 얼굴에 가상의 객체를 합성하기 위한 정보로 이용한다. 다양한 실험을 통하여 사용자 얼굴의 움직임에 대한 정보 추출의 정확도를 검증하였다.

• 천문학회보
• /
• 제42권2호
• /
• pp.39.2-39.2
• /
• 2017

Over the last decade, deep studies of nearby galaxies have led to the discovery of vast stellar envelopes that are often rich in substructure. These components are naturally predicted in models of hierarchical galaxy assembly, and their observed properties place important constraints on the amount, nature, and history of satellite accretion. One of the most effective ways of mapping the peripheral regions of galaxies is through resolved star studies. Using wide-field cameras equipped to 8 m class telescopes, it has recently become possible to extend these studies to systems beyond the Local Group. Located at a distance of 3.6 Mpc, M81 is a prime target for wide-field mapping of its resolved stellar content. In this talk, we present the detailed results from our deep wide-field imaging survey of the M81 group with the Hyper Suprime-Cam (HSC), on the Subaru Telescope. We report on the analysis of the structures, stellar populations, and metallicities of old dwarf galaxies such as NGC3077, IKN, KDG061, as well as young stellar systems such as Arp's Loop and Holmberg IX. Several candidates for yet-undiscovered faint dwarf galaxies and young stellar clumps in the M81 group will also be introduced. The peculiar galaxy NGC3077 has been classified as the irregular galaxy. Okamoto et al. (2015, ApJ 809, L1) discovered an extended halo structure with S-shape elongated tails, obvious feature of tidal interaction. With a help of numerical simulation by Penarrubia et al. (2009, ApJ 698, 222), we will demonstrate that this tidal feature was formed during the latest close encounters between M81, M82, and NGC 3077, which induced star formation in tidally stripped gas far from the main bodies of galaxies. It is not clear whether the latest tidal interaction was the first close encounters of three galaxies. If NGC3077 is still surrounded by the dark matter halo, it implies that NGC3077 has undergone the first tidal stripping by larger companions. Kinematic studies of inter galactic globular clusters and planetary nebulae would tell us the past history of tidal interaction in this group of galaxies.

• 한국정보통신학회논문지
• /
• 제22권1호
• /
• pp.125-130
• /
• 2018

선형판별분석(LDA) 기법은 특징벡터의 차원을 줄이거나 클래스 식별에 이용되는 통계적 분석 방법이다. 그러나 선형 분리가 불가능한 데이터 집합의 경우에는 비선형 함수를 이용하여 특징벡터를 고차원의 공간으로 사상(mapping) 시켜줌으로써 선형 분리가 가능하도록 만들 수 있는데, 이러한 기법을 일반화된 판별분석(GDA) 또는 커널판별분석(KDA) 기법이라고 한다. 본 연구에서는 인터넷에 공개되어 있는 능동소나 표적신호에 LDA 및 GDA 기법을 이용하여 표적식별 실험을 수행하고, 그 결과를 비교/분석하였다. 실험 결과 104개의 테스트 데이터에 대해 LDA 기법으로는 73.08% 인식률을 얻었으나 GDA 기법으로는 95.19%로 기존의 MLP 또는 커널 기반 SVM에 비해 나은 성능을 보였다.