• 정보처리학회논문지B
• /
• 제17B권5호
• /
• pp.375-380
• /
• 2010

객체 윤곽을 추출하는 대부분의 기존 방법들은 단일객체의 윤곽검출에 대해 연구하였다. 그러나, 실 세계에서는 복수객체가 일반적이다. 본 논문에서 제안한 복수객체 윤곽추출 알고리즘은 2 단계로 구성된다. 첫 번째 단계는 외적 및 내적을 사용하여 초기에 복수객체를 포함한 윤곽을 고속으로 분리하고 연결하여 각각이 윤곽이 단일 개체만을 포함하는 방법을 제안한다. 두 번째 단계는 각각의 윤곽에 포함된 단일 객체의 윤곽을 추출하는 개선된 능동윤곽모델 알고리즘을 설명한다. 여러 실험영상에 대한 실험결과는 다른 방법과 비교하여 속도가 빠르며 정확하게 윤곽을 추출한다.

• 해양환경안전학회지
• /
• 제29권2호
• /
• pp.139-146
• /
• 2023

양식장 부표 등과 같은 해상의 소형 장애물을 탐지하고 거리와 방위를 시각화시켜 주는 해상물체탐지시스템은 선체운동으로 인한 오차를 보정하기 위해 3축 짐벌이 장착되어 있지만, 파도 등에 의한 카메라와 해상물체의 상하운동으로 발생하는 거리오차를 보정하지 못하는 한계가 있다. 이에 본 연구에서는 외부환경에 따른 수면의 움직임으로 발생하는 해상물체탐지시스템의 거리오차를 분석하고, 이를 평균필터와 이동평균필터로 보정하고자 한다. 가우시안 표준정규분포를 따르는 난수를 이미지 좌표에 가감하여 불규칙파에 의한 부표의 상승 또는 하강을 재현하였다. 이미지 좌표의 변화에 따른 계산거리, 평균필터와 이동평균필터를 통한 예측거리 그리고 레이저 거리측정기에 의한 실측거리를 비교하였다. phase 1,2에서 불규칙파에 의한 이미지 좌표의 변화로 오차율이 최대 98.5%로 증가하였지만, 이동평균필터를 사용함으로써 오차율은 16.3%로 감소하였다. 오차보정 능력은 평균필터가 더 좋았지만 거리변화에 반응하지 못하는 한계가 있었다. 따라서 해상물체탐지시스템 거리오차 보정을 위해 이동평균필터를 사용함으로써 실시간 거리변화에 반응하고 오차율을 크게 개선할 수 있을 것으로 판단된다.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 1995년도 학술대회
• /
• pp.125-129
• /
• 1995

Motion blur, which occurs when relative motion between a camera and an object exits, is inevitable in moving pictures. By this reason there have been many research results for reducing such motion blur, especially for spatially variant case. Int he present paper, an adaptive iterative restoration method is applied to reduce the irregular motion blur in the image.

• 한국정밀공학회지
• /
• 제16권11호
• /
• pp.89-97
• /
• 1999

A new parallel gripper mechanism is proposed in this work. This device has a parallelogrammic platform which can be flexibly folded. Therefore, this mechanism not only can be used to grasp an object having irregular shape or large volume, but also can be used as a micro-positioning device after grasping. Based on the position and kinematic analysis for this mechanism, this mechanism has been developed and the motion performance has been tested to corroborate the effectiveness of this mechanism.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제54권10호
• /
• pp.604-614
• /
• 2005

Color segmentation takes great attentions since a color is an effective and robust visual cue for characterizing one object from other objects. Color segmentation is, however, suffered from color variation induced from irregular illumination changes. This paper proposes a reliable color modeling approach in HSI (Hue-Saturation-Intensity) rotor space considering intensity information by adopting B-spline curve fitting to make a mathematical model for statistical characteristics of a color with respect to brightness. It is based on the fact that color distribution of a single-colored object is not invariant with respect to brightness variations even in HS (Hue-Saturation) plane. The proposed approach is applied for the segmentation of human skin areas successfully under various illumination conditions.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권6호
• /
• pp.2217-2229
• /
• 2015

Since the variations of illumination, the irregular changes of human shapes, and the partial occlusions, multiple person tracking is a challenging work in computer vision. In this paper, we propose a graph clustering method based on spatio-temporal information of moving objects for multiple person tracking. First, the part-based model is utilized to localize individual foreground regions in each frame. Then, we heuristically leverage the spatio-temporal constraints to generate a set of reliable tracklets. Finally, the graph shift method is applied to handle tracklet association problem and consequently generate the completed trajectory for individual object. The extensive comparison experiments demonstrate the superiority of the proposed method.

• 한국군사과학기술학회지
• /
• 제5권1호
• /
• pp.37-44
• /
• 2002

An echo signal in the active sonar using a high frequency is mainly formed of a specular reflection from the surface of an object along with several equivalent scatter inside, which are characterized by the spatial distribution of the highlights on the object. This thesis proposed a model in which the synthesized echo signal can be expressed as a distributed simulated target. The proposed model is obtained after composing a signal based on the movement of highlights relative to the aspect angle from the discontinuous point of an external hull with a strong reflection from a spheroid underwater target. Because the proposed algorithm includes a synthesis of the signals related to the highlight spacial distribution, it can be applied to all kinds of systems used at a short range, and similar results were obtained to the actual measured results of all reflected signals in previous literature referring to the irregular factor application of an envelope.

• 정보처리학회논문지C
• /
• 제13C권6호
• /
• pp.669-676
• /
• 2006

본 논문에서는 실시간 객체모델인 Time-triggered Message-triggered Object (TMO)를 사용하여 분산 실시간 멀티미디어 스트리밍 서비스를 효과적으로 개발 할 수 있는 새로운 소프트웨어 프레임워크와 스트림 동기화 메커니즘을 소개한다. 본 프레임워크의 목적은 개발자로 하여금 복잡한 실시간 멀티미디어 스트리밍 서비스를 쉽게 설계하고 적시 스트리밍 기능들을 제공하는데 있다. 이를 위해서 본 프레임워크는 Multimedia Streaming TMO, MMStream TMO Support Library 그리고 TMO Support Middleware로 구성된다. 특히, MMStream TMO와 동기화 기법이 제공하는 time-trigger번 특성은 QoS 보장이 어려운 통신 채널과 시스템 환경에서 불규칙한 미디어 데이터 전달 및 처리하는 역할을 실시간으로 수행한다. 결론적으로 본 논문에서 제시한 프레임워크의 적시 서비스 능력은 향후 실시간 멀티미디어 스트리밍 서비스 개발에 기여할 것으로 기대된다.

• 대한안전경영과학회지
• /
• 제22권1호
• /
• pp.1-8
• /
• 2020

Maintenance of power distribution facilities is a significant subject in the power supplies. Fault caused by deterioration in power distribution facilities may damage the entire power distribution system. However, current methods of diagnosing power distribution facilities have been manually diagnosed by the human inspector, resulting in continuous pole accidents. In order to improve the existing diagnostic methods, a thermal image analysis model is proposed in this work. Using a thermal image technique in diagnosis field is emerging in the various engineering field due to its non-contact, safe, and highly reliable energy detection technology. Deep learning object detection algorithms are trained with thermal images of a power distribution facility in order to automatically analyze its irregular energy status, hereby efficiently preventing fault of the system. The detected object is diagnosed through a thermal intensity area analysis. The proposed model in this work resulted 82% of accuracy of detecting an actual distribution system by analyzing more than 16,000 images of its thermal images.

• Journal of Information Processing Systems
• /
• 제17권4호
• /
• pp.787-800
• /
• 2021

Three-dimensional (3D) object classification tasks using point clouds are widely used in 3D modeling, face recognition, and robotic missions. However, processing raw point clouds directly is problematic for a traditional convolutional network due to the irregular data format of point clouds. This paper proposes a pointwise convolution neural network (CNN) structure that can process point cloud data directly without preprocessing. First, a 2D convolutional layer is introduced to percept coordinate information of each point. Then, multiple 2D convolutional layers and a global max pooling layer are applied to extract global features. Finally, based on the extracted features, fully connected layers predict the class labels of objects. We evaluated the proposed pointwise CNN structure on the ModelNet10 dataset. The proposed structure obtained higher accuracy compared to the existing methods. Experiments using the ModelNet10 dataset also prove that the difference in the point number of point clouds does not significantly influence on the proposed pointwise CNN structure.