• 한국멀티미디어학회논문지
• /
• 제24권5호
• /
• pp.659-666
• /
• 2021

With the continuous development of deep learning, human behavior recognition algorithms have achieved good results. However, in a multi-person recognition environment, the complex behavior environment poses a great challenge to the efficiency of recognition. To this end, this paper proposes a multi-person pose estimation model. First of all, the human detectors in the top-down framework mostly use the two-stage target detection model, which runs slow down. The single-stage YOLOv3 target detection model is used to effectively improve the running speed and the generalization of the model. Depth separable convolution, which further improves the speed of target detection and improves the model's ability to extract target proposed regions; Secondly, based on the feature pyramid network combined with context semantic information in the pose estimation model, the OHEM algorithm is used to solve difficult key point detection problems, and the accuracy of multi-person pose estimation is improved; Finally, the Euclidean distance is used to calculate the spatial distance between key points, to determine the similarity of postures in the frame, and to eliminate redundant postures.

• 제어로봇시스템학회논문지
• /
• 제17권9호
• /
• pp.916-921
• /
• 2011

Mobile robot localization is the task of estimating the robot pose in a given environment. This research deals with outdoor localization based on an elevation map. Since outdoor environments are large and contain many complex objects, it is difficult to robustly estimate the robot pose. This paper proposes a Hausdorff distance-based map matching method. The Hausdorff distance is exploited to measure the similarity between extracted features obtained from the robot and elevation map. The experiments and simulations show that the proposed Hausdorff distance-based map matching is useful for robust outdoor localization using an elevation map. Also, it can be easily applied to other probabilistic approaches such as a Markov localization method.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2002년도 하계종합학술대회 논문집(3)
• /
• pp.227-230
• /
• 2002

In this paper we concentrate on overall direction of the gaze based on a head pose for human computer interaction. To decide a gaze direction of user in a image, it is important to pick up facial feature exactly. For this, we binarize the input image and search two eyes and the mouth through the similarity of each block ( aspect ratio, size, and average gray value ) and geometric information of face at the binarized image. We create a imaginary plane on the line made by features of the real face and the pin hole of the camera to decide the head orientation. We call it the virtual facial plane. The position of a virtual facial plane is estimated through projected facial feature on the image plane. We find a gaze direction using the surface normal vector of the virtual facial plane. This study using popular PC camera will contribute practical usage of gaze tracking technology.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 ITC-CSCC -1
• /
• pp.11-14
• /
• 2000

In this paper, we propose a fast and practical head pose estimation scheme fur eye-head controlled human computer interface with non-constrained background. The method we propose uses complete graph matching from thresholded images and the two blocks showing the greatest similarity are selected as eyes, we also locate mouth and nostrils in turn using the eye location information and size information. The average computing time of the image(360*240) is within 0.2(sec) and we employ template matching method using angles between facial features for head pose estimation. It has been tested on several sequential facial images with different illuminating conditions and varied head poses, It returned quite a satisfactory performance in both speed and accuracy.

• 통합자연과학논문집
• /
• 제3권3호
• /
• pp.143-147
• /
• 2010

Since it is of great importance to know how a ligand binds to a receptor, there have been a lot of efforts to improve the quality of prediction of docking poses. Earlier efforts were focused on improving search algorithm and scoring function in a docking program resulting in a partial improvement with a lot of variations. Although these are basically very important and essential, more tangible improvements came from the reduction of search space. In a normal docking study, the approximate active site is assumed to be known. After defining active site, scoring functions and search algorithms are used to locate the expected binding pose within this search space. A good search algorithm will sample wisely toward the correct binding pose. By careful study of receptor structure, it was possible to prioritize sub-space in the active site using "receptor-based pharmacophores" or "hot spots". In a sense, these techniques reduce the search space from the beginning. Further improvements were made when the bound ligand structure is available, i.e., the searching could be directed by molecular similarity using ligand information. This could be very helpful to increase the accuracy of binding pose. In addition, if the biological activity data is available, docking program could be improved to the level of being useful in affinity prediction for a series of congeneric ligands. Since the number of co-crystal structures is increasing in protein databank, "Ligand-Guided Docking" to reduce the search space would be more important to improve the accuracy of docking pose prediction and the efficiency of virtual screening. Further improvements in this area would be useful to produce more reliable docking programs.

• IEIE Transactions on Smart Processing and Computing
• /
• 제6권4호
• /
• pp.262-268
• /
• 2017

Intelligent video surveillance systems have been developed to monitor global areas and find specific target objects using a large-scale database. However, person re-identification presents some challenges, such as pose change and occlusions. To solve the problems, this paper presents an improved person re-identification method using sparse representation and saliency-based dictionary construction. The proposed method consists of three parts: i) feature description based on salient colors and textures for dictionary elements, ii) orthogonal atom selection using cosine similarity to deal with pose and viewpoint change, and iii) measurement of reconstruction error to rank the gallery corresponding a probe object. The proposed method provides good performance, since robust descriptors used as a dictionary atom are generated by weighting some salient features, and dictionary atoms are selected by reducing excessive redundancy causing low accuracy. Therefore, the proposed method can be applied in a large scale-database surveillance system to search for a specific object.

• 정보처리학회논문지A
• /
• 제10A권4호
• /
• pp.357-368
• /
• 2003

3차원 모델의 형상 유사성 평가는 의학, 기계 공학, 분자 생물학 등의 많은 분야에서 매우 중요하다. 더욱이 3차원 모델이 웹 상에 보편화됨에 따라 3차원 모델들의 분류와 검색에 관한 연구들이 활발하게 이루어지고 있다. 본 논문에서는 3차원 형상 표현 방법들과 유사성 평가에 대한 주요 개념들을 기술하고, 최근의 형상 비교에 관한 연구들을 다해상도, 위상 기하학, 2차원 영상, 통계학 기반 방법들로 분류하여 그 특징들을 분석하였다. 또한 논문에서 채택한 유일성, 강인성, 불변성, 다해상도, 효율성, 비교범위와 같은 기준을 사용하여 그 성능을 비교 평가하였다. 다해상도 기반 방법은 비교를 위한 계산 시간은 감소시킨 반면 전처리 시간은 증가시켰다. 기하 및 위상 정보를 이용한 방법은 보다 다양한 형태의 모델들을 비교할 수 있었고 부분적인 형상 비교에도 강인하였다. 2차원 영상을 이용한 방법들은 시간 및 공간 복잡도가 높게 나타났다. 통계학 기반 방법들은 포즈 정규화 작업 없이 형상 비교가 가능하였고, 어파인 변환 및 잡음에도 강인한 결과를 보였다.

• 한국정보과학회논문지:데이타베이스
• /
• 제35권5호
• /
• pp.379-390
• /
• 2008

영상 데이타의 증가로 인해 여러 분야에 걸쳐 유사 영상 검색에 대한 수요는 계속해서 증가하고 있으며, 좀 더 정확한 유사 영상 검색을 위해 내용기반영상검색 기술 분야에 대한 많은 연구가 이루어지고 있다. 내용기반영상검색에서는 유사한 영상을 검색하기 위해 색상, 모양, 질감 등의 영상 자체의 내용을 사용한다. 하지만 위성 사진과 같이 색상 정보의 사용이 어렵고 영상 자체가 매우 복잡한 영상의 경우는 내용기반영상검색 기술을 이용해도 유사성 검색이 쉽지 않다. 그 이유는 색상 정보의 사용이 어려워 모양 정보를 이용하려면 영상 세그멘테이션을 통해 물체의 모양을 분리해내야 하는데 영상이 매우 복잡한 경우 영상 세그멘테이션의 결과가 좋지 않기 때문이다. 따라서 본 논문에서는 영상 세그멘테이션을 사용하지 않고 유사성 검색을 하는 방법을 제안하여 유사 위성 영상을 검색하고자 한다. 이를 위해 영상 세그멘테이션을 사용하지 않는 모양 기술자인 SIFT keypoint descriptor를 기반으로 영상의 유사도를 정의하여 유사성 검색을 한다. 실험을 통해 제안한 방법을 이용함으로써 색상 정보의 사용이 어렵고 영상 자체가 매우 복잡한 위성 영상의 유사성 검색의 정확성을 향상 시킬 수 있음을 보인다.

• The Journal of Asian Finance, Economics and Business
• /
• 제6권4호
• /
• pp.179-187
• /
• 2019

This study aims to investigate the effects of market commonality and resource similarity on price competition and the recursive consequences in the Korean lodging market. Price comparison among hotels in the same geographic market has been facilitated through the development of information technology, rendering little search cost of consumers. While the literature implies the heterogeneous price attack and response among hotels, a limited number of empirical researches focus on the asymmetric and recursive pattern in the competitive dynamics. This study empirically examines the price interactions in the Korean lodging market based on the theoretical framework of competitive price interactions and countervailing power. Demonstrating superiority to the spatial lag model and the ordinary least squares in the estimation, the results from spatial error model suggest that the hotels with longer operational history pose an asymmetric impact on the price of the newer hotels. The asymmetry is also found in chain hotels over the independent, further implying the possibility of predatory pricing. The findings of this study provide the evidence of a hierarchical structure in the price competition, with different countervailing power by the resources of the hotels. Theoretical and managerial implications are discussed, with suggestions for future study.

• Bulletin of the Korean Chemical Society
• /
• 제34권4호
• /
• pp.1212-1220
• /
• 2013

The human coagulation factor XI (FXI) is a serine protease that plays a significant role in blocking of the blood coagulation cascade as an attractive antithrombotic target. Selective inhibition of FXIa (an activated form of factor XI) disrupts the intrinsic coagulation pathway without affecting the extrinsic pathway or other coagulation factors such as FXa, FIIa, FVIIa. Furthermore, targeting the FXIa might significantly reduce the bleeding side effects and improve the safety index. This paper reports on a docking-based three dimensional quantitative structure activity relationship (3D-QSAR) study of the potent FXIa inhibitors, the chloro-phenyl tetrazole scaffold series, using comparative molecular field analysis (CoMFA) and comparative molecular similarity analysis (CoMSIA) methods. Due to the characterization of FXIa binding site, we classified the alignment of the known FXIa inhibitors into two groups according to the docked pose: S1-S2-S4 and S1-S1'-S2'. Consequently, highly predictive 3D-QSAR models of our result will provide insight for designing new potent FXIa inhibitors.