• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2007년도 Proceedings of ISRS 2007
• /
• pp.234-237
• /
• 2007

Road networks are important geospatial databases for various GIS (Geographic Information System) applications. Road digital maps may contain geometric spatial errors due to human and scanning errors, but manually updating roads information is time consuming. In this paper, we developed a new road features updating methodology using from multispectral high-resolution satellite image and pre-existing vector map. The approach is based on initial seed point generation using line segment matching and a modified double snake algorithm. Firstly, we conducted line segment matching between the road vector data and the edges of image obtained by Canny operator. Then, the translated road data was used to initialize the seed points of the double snake model in order to refine the updating of road features. The double snake algorithm is composed of two open snake models which are evolving jointly to keep a parallel between them. In the proposed algorithm, a new energy term was added which behaved as a constraint. It forced the snake nodes not to be out of potential road pixels in multispectral image. The experiment was accomplished using a QuickBird pan-sharpened multispectral image and 1:5,000 digital road maps of Daejeon. We showed the feasibility of the approach by presenting results in this urban area.

• 대한전자공학회논문지
• /
• 제27권11호
• /
• pp.32-41
• /
• 1990

본 논문에서는 시각정보에 의하여 로봇을 제어하기위해 영상으로부터 추출되는 feature를 이용하는 한 방법을 제안한다. 특별히 feature에 대한 수학적인 정의를 제안하였으며 로봇의 움직임과 feature vector의 미소한 변화 사이의 관계를 기술하였다. 이 과정에서 feature jacobian matrix와 그의 gene-ralized inverse가 사용되었다. 로봇 자유도의 수보다 많은 feature를 사용하면 visual servoing의 성능을 향상시킬 수 있었다. 여러 예를 통하여, 본 논문에서 제안된 방법이 유효함을 보였다.

• 한국멀티미디어학회논문지
• /
• 제22권8호
• /
• pp.832-843
• /
• 2019

Prostate cancer is a high-risk with a high incidence and is a disease that occurs only in men. Accurate diagnosis of cancer is necessary as the incidence of cancer patients is increasing. Prostate cancer is also a disease that is difficult to predict progress, so it is necessary to predict in advance through prognosis. Therefore, in this paper, grade classification is attempted based on texture feature extraction. There are two main methods of classification: Uses One-way Analysis of Variance (ANOVA) to determine whether texture features are significant values, compares them with all texture features and then uses only one classification i.e. Benign versus. The second method consisted of more detailed classifications without using ANOVA for better analysis between different grades. Results of both these methods are compared and analyzed through the machine learning models such as Support Vector Machine and K-Nearest Neighbor. The accuracy of Benign versus Grade 4&5 using the second method with the best results was 90.0 percentage.

• 정보보호학회논문지
• /
• 제20권5호
• /
• pp.133-138
• /
• 2010

본 논문은 얼굴과 음성 정보를 사용한 다중 바이오 인증에서, 특정 단계의 융합과 결정 단계의 융합을 동시에 수행하는 다단계 융합 방법을 제안한다. 얼굴과 음성 특징을 1차 융합한 얼굴 음성 융합특징에 대해 Support Vector Machines(SVM)을 생성한 후, 이 융합특징 SVM 인증기의 결정과 얼굴 SVM 인증기의 결정, 음성 SVM 인증기의 결정들을 다시 2차 융합하여 최종 인증 여부를 결정한다. XM2VTS 멀티모달 데이터베이스를 사용하여 특징 단계 융합, 결정 단계 융합, 다단계 융합 인증을 비교 실험한 결과, 제안한 다단계 융합에 의한 인증이 가장 우수한 성능을 보였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권10호
• /
• pp.3211-3229
• /
• 2022

To solve the problems existing in the process of Weibo disaster rumor recognition, such as lack of corpus, poor text standardization, difficult to learn semantic information, and simple semantic features of disaster rumor text, this paper takes Sina Weibo as the data source, constructs a dataset for Weibo disaster rumor recognition, and proposes a deep learning model BERT_AT_Stacked LSTM for Weibo disaster rumor recognition. First, add adversarial disturbance to the embedding vector of each word to generate adversarial samples to enhance the features of rumor text, and carry out adversarial training to solve the problem that the text features of disaster rumors are relatively single. Second, the BERT part obtains the word-level semantic information of each Weibo text and generates a hidden vector containing sentence-level feature information. Finally, the hidden complex semantic information of poorly-regulated Weibo texts is learned using a Stacked Long Short-Term Memory (Stacked LSTM) structure. The experimental results show that, compared with other comparative models, the model in this paper has more advantages in recognizing disaster rumors on Weibo, with an F1_Socre of 97.48%, and has been tested on an open general domain dataset, with an F1_Score of 94.59%, indicating that the model has better generalization.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.1032-1035
• /
• 2003

The field of handwriting recognition has been researched for many years. A hybrid classifier has been proven to be able to increase the recognition rate compared with a single classifier. In this paper, we combine support vector machine (SVM) and hidden Markov model (HMM) for offline handwritten numeral recognition. To improve the performance, we extract features adapted for each classifier and propose the modified SVM decision structure. The experimental results show that the proposed method can achieve improved recognition rate for handwritten numeral recognition.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 하계종합학술대회 논문집(3)
• /
• pp.177-180
• /
• 2000

Recently, a feature extraction method based on decision boundary has been proposed for neural networks. The method is based on the fact that all the features necessary to achieve the same classification accuracy as in the original space can be obtained from the vectors normal to decision boundaries. However, the normal vector was estimated numerically. resulting in inaccurate estimation and a long computational time. In this paper. we propose a new method to calculate the normal vector analytically. Experiments show that the proposed method provides a better performance.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권12호
• /
• pp.6190-6213
• /
• 2019

For small-scale vector data, restrictions on watermark scheme capacity and robustness limit the use of copyright protection. A watermarking scheme based on robust geometric features and capacity maximization strategy that simultaneously improves capacity and robustness is presented in this paper. The distance ratio and angle of adjacent vertices are chosen as the watermark domain due to their resistance to vertex and geometric attacks. Regarding watermark embedding and extraction, a capacity-improved strategy based on quantization index modulation, which divides more intervals to carry sufficient watermark bits, is proposed. By considering the error tolerance of the vector map and the numerical accuracy, the optimization of the capacity-improved strategy is studied to maximize the embedded watermark bits for each vertex. The experimental results demonstrated that the map distortion caused by watermarks is small and much lower than the map tolerance. Additionally, the proposed scheme can embed a copyright image of 1024 bits into vector data of 150 vertices, which reaches capacity at approximately 14 bits/vertex, and shows prominent robustness against vertex and geometric attacks for small-scale vector data.

• 한국통신학회논문지
• /
• 제29권4C호
• /
• pp.494-504
• /
• 2004

사용자의 시선 위치를 파악하는 연구는 많은 응용분야를 가지고 지난 몇년간 눈부시게 발전되어 왔다. 기존의 대부분 연구에서는 영상 처리 방법만에 의존하여 시선 위치 추적 연구를 수행하였기 때문에 처리 속도도 늦고 많은 사용 제약을 가지는 문제점이 있었다. 이 논문에서는 적외선 조명이 부착된 단일 카메라를 이용한 컴퓨터 비전 시스템으로 시선 위치 추적 연구를 수행하였다. 사용자의 시선 위치를 파악하기 위해서는 얼굴 특징점의 위치를 추적해야하는데, 이를 위하여 이 논문에서는 적의선 기반 카메라와 SVM(Support Vector Machine) 알고리즘을 사용하였다. 사용자가 모니터상의 임의의 지점을 쳐다볼 때 얼굴 특징점의 3차원 위치는 3차원 움직임량 추정(3D motion estimation) 및 아핀 변환(affine transformation)에 의해 계산되어 질 수 있다. 얼굴 특징점의 변화된 3차원 위치가 계산되면. 이로부터 3개 이상의 얼굴 특징점으로부터 생성되는 얼굴 평면 및 얼굴 평면의 법선 벡터가 구해지게 되며, 이러한 법선 백터가 모니터 스크린과 만나는 위치가 사용자의 시선위치가 된다. 또한. 이 논문에서는 보다 정확한 시선 위치를 파악하기 위하여 사용자의 눈동자 움직임을 추적하였으며 이를 위하여 신경망(다층 퍼셉트론)을 사용하였다. 실험 결과, 얼굴 및 눈동자 움직임에 의한 모니터상의 시선 위치 정확도는 약 4.2cm의 최소 자승 에러성능을 나타냈다.

• 한국멀티미디어학회논문지
• /
• 제21권4호
• /
• pp.478-484
• /
• 2018

The choice of the motion features influences the result of the human action recognition method directly. Many factors often influence the single feature differently, such as appearance of the human body, environment and video camera. So the accuracy of action recognition is restricted. On the bases of studying the representation and recognition of human actions, and giving fully consideration to the advantages and disadvantages of different features, the Dense Trajectories(DT) algorithm is a very classic algorithm in the field of behavior recognition feature extraction, but there are some defects in the use of optical flow images. In this paper, we will use the improved Dense Trajectories(iDT) algorithm to optimize and extract the optical flow features in the movement of human action, then we will combined with Support Vector Machine methods to identify human behavior, and use the image in the KTH database for training and testing.