• 한국게임학회 논문지
• /
• 제11권2호
• /
• pp.161-166
• /
• 2011

본 논문에서는 동작인식 위한 정확한 배경 분할 및 특징점 추출 방법을 제안한다. 배경 분할 과정에서는 먼저, HSV 입력 이미지를 RGB 색상 공간에서 HSV 색상 공간으로 변환한 뒤, H와 S 값에 대한 두 개의 임계치를 사용하여 살색 영역을 분할, 프레임간의 차영상을 이용하여 움직임이 있는 영역을 추출한다. 차영상에서 발생하는 잔상 영역을 제거하기 위하여 헤시안 어파인 영역 검출기를 적용하고, 잡음이 제거된 차 영상과 살색 영역의 이진화 영상을 이용하여 사람의 동작이 나타나는 영역을 분할한다. 특징점 추출 과정은 전체 영상을 블록 단위로 나눠서 각 블록 안에서 분할된 영상에 포함되는 픽셀들의 중점을 구하여 특징점을 추출한다. 실험결과 복잡한 환경에서도 정확한 배경 분할과 사용자 동작을 대표하는 특징점 추출이 약 12 fps로 가능함을 알 수 있었다.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 2003년도 ISIS 2003
• /
• pp.632-635
• /
• 2003

This paper describes a system fur tracking multiple faces in an input video sequence using facial convex hull based facial segmentation and robust hausdorff distance. The algorithm adapts skin color reference map in YCbCr color space and hair color reference map in RGB color space for classifying face region. Then, we obtain an initial face model with preprocessing and convex hull. For tracking, this algorithm computes displacement of the point set between frames using a robust hausdorff distance and the best possible displacement is selected. Finally, the initial face model is updated using the displacement. We provide an example to illustrate the proposed tracking algorithm, which efficiently tracks rotating and zooming faces as well as existing multiple faces in video sequences obtained from CCD camera.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제3권1호
• /
• pp.87-92
• /
• 2003

This paper describes a system for tracking multiple faces in an input video sequence using facial convex hull based facial segmentation and robust hausdorff distance. The algorithm adapts skin color reference map in YCbCr color space and hair color reference map in RGB color space for classifying face region. Then, we obtain an initial face model with preprocessing and convex hull. For tracking, this algorithm computes displacement of the point set between frames using a robust hausdorff distance and the best possible displacement is selected. Finally, the initial face model is updated using the displacement. We provide an example to illustrate the proposed tracking algorithm, which efficiently tracks rotating and zooming faces as well as existing multiple faces in video sequences obtained from CCD camera.

• 한국콘텐츠학회논문지
• /
• 제9권5호
• /
• pp.1-12
• /
• 2009

올바른 피부색 검출은 사람의 얼굴 검출 및 동작 분석에서 매우 중요한 전처리과정에 속한다. 피부 검출은 일반적으로 화소의 칼라 공간을 Non-RGB로 변형하고, 피부색의 조명 요소를 제거한 다음 피부색 분포 모델에 의해 Skin과 Non-Skin으로 분류하는 3단계로 진행된다. 이는 피부색 검출이 칼라 공간, 조명 요소의 존재 여부, 피부 모델링 방법에 따라 수행 성능에 많은 영향을 받기 때문이다. 본 연구에서는 조명 조건에 따라 피부색 모델의 범위에 차이가 있다는 사실에 기초하여 다양한 조명 조건과 복잡한 배경을 가진 영상에서 효과적으로 다인종의 피부색을 분류해내 기 위한 3차원 피부색 모델을 제시하고자 한다. 제안된 피부색 모델은 화소의 칼라 공간을 YCbCr공간으로 변형하고, 각 요소(Y, Cb, Cr) 값에 의한 3차원 피부색 모델을 형성한다. 다인종의 피부색을 함께 분할하기 위해 인종(백인, 흑인, 황인)별 피부색 모델을 먼저 생성한 후 각각의 모델에서 피부색 확률에 따라 결합한 다인종을 위한 통합 모델을 생성하였다. 또한 우리는 적은 양의 훈련 데이터로 피부색 영역을 올바르게 검출할 수 있도록 여러 단계의 피부색 영역을 설정하였다.

• 전자공학회논문지CI
• /
• 제47권2호
• /
• pp.11-20
• /
• 2010

본 논문에서는 피부색 정보를 이용한 얼굴 후보 검출 방법과 얼굴의 구조적 특징을 이용한 얼굴 확인 방법으로 구성된 얼굴 검출 시스템을 제안한다. 먼저 제안하는 얼굴 후보 검출 방법은 피부색 영역과 피부색의 주변 영역에 대한 이미지 분할과 병합 알고리듬을 이용한다. 이미지 분할과 병합 알고리듬의 적용은 복잡한 이미지에 존재하는 다양한 얼굴들을 후보로 검출할 수 있다. 그리고 제안하는 얼굴 확인 방법은 얼굴을 지역적인 특징에 따라 분류 가능한 부분 얼굴 분류기를 사용하여 얼굴의 구조적 특징을 판단하고, 얼굴과 비-얼굴을 구별한다. 부분 얼굴 분류기는 학습 과정에서 얼굴 이미지만을 사용하고, 비-얼굴 이미지는 고려하지 않기 때문에 적은 수의 훈련 이미지를 사용한다. 실험 결과 제안한 얼굴 후보 검출 방법은 기존의 방법보다 평균 9.55% 많은 얼굴을 후보로 검출하였다. 그리고 얼굴/비-얼굴 분류 실험에서 비-얼굴에 대한 분류율이 99%일 때 기존의 분류기보다 평균 4.97% 높은 얼굴 분류율을 달성 하였다.

• 로봇학회논문지
• /
• 제15권4호
• /
• pp.316-322
• /
• 2020

Endoscopic spine surgery is an advanced surgical technique for spinal surgery since it minimizes skin incision, muscle damage, and blood loss compared to open surgery. It requires, however, accurate positioning of an endoscope to avoid spinal nerves and to locate the endoscope near the target disk. Before the insertion of the endoscope, a guide needle is inserted to guide it. Also, the result of the surgery highly depends on the surgeons' experience and the patients' CT or MRI images. Thus, for the training, a number of haptic simulators for spinal needle insertion have been developed. But, still, it is difficult to be used in the medical field practically because previous studies require manual segmentation of vertebrae from CT images, and interaction force between the needle and soft tissue has not been considered carefully. This paper proposes AI-based automatic vertebrae CT-image segmentation and haptic rendering method using the proposed need-tissue interaction model. For the segmentation, U-net structure was implemented and the accuracy was 93% in pixel and 88% in IoU. The needle-tissue interaction model including puncture force and friction force was implemented for haptic rendering in the proposed spinal needle insertion simulator.

• 한국소음진동공학회:학술대회논문집
• /
• 한국소음진동공학회 2013년도 추계학술대회 논문집
• /
• pp.150-151
• /
• 2013

• 한국멀티미디어학회논문지
• /
• 제25권8호
• /
• pp.1212-1223
• /
• 2022

Skin lesions are common diseases that range from skin rashes to skin cancer, which can lead to death. Note that early diagnosis of skin diseases can be important because early diagnosis of skin diseases considerably can reduce the course of treatment and the harmful effect of the disease. Recently, the development of computer-aided diagnosis (CAD) systems based on artificial intelligence has been actively made for the early diagnosis of skin diseases. In a typical CAD system, the accurate classification of skin lesion types is of great importance for improving the diagnosis performance. Motivated by this, we propose a novel deep ensemble classification with multi-scale attention networks. The proposed deep ensemble networks are jointly trained using a single loss function in an end-to-end manner. In addition, the proposed deep ensemble network is equipped with a multi-scale attention mechanism and segmentation information of the original skin input image, which improves the classification performance. To demonstrate our method, the publicly available human skin disease dataset (HAM 10000) and the private animal skin lesion dataset were used for the evaluation. Experiment results showed that the proposed methods can achieve 97.8% and 81% accuracy on each HAM10000 and animal skin lesion dataset. This research work would be useful for developing a more reliable CAD system which helps doctors early diagnose skin diseases.

• 한국CDE학회논문집
• /
• 제19권2호
• /
• pp.111-118
• /
• 2014

In tangible augmented reality (AR) environments, the user interacts with virtual objects by manipulating their physical counterparts, but he or she often encounters awkward situations in which his or her hands are occluded by the augmented virtual objects, which causes great difficulty in figuring out hand positions, and reduces both immersion and ease of interaction. To solve the problem of such hand occlusion, skin color information has been usefully exploited. In this paper, we propose an approach to simple and effective construction of a skin color map which is suitable for hand segmentation and tangible AR interaction. The basic idea used herein is to obtain hand images used in a target AR environment by simple image subtraction and to represent their color information by a convex polygonal map in the YCbCr color space. We experimentally found that the convex polygonal map is more accurate in representing skin color than a conventional rectangular map. After implementing a solution for resolving hand occlusion using the proposed skin color map construction, we showed its usefulness by applying it to virtual design evaluation of digital handheld products in a tangible AR environment.

• 한국생활과학회지
• /
• 제24권4호
• /
• pp.555-569
• /
• 2015

During a design process of a protective equipment for sports activities, minimizing movement restrictions is important for enhancing its functions particularly for protection. This study presents a three-dimensional(3D) modeling methodology for designing baseball catcher's leg guards that will allow maximum possible performance, while providing necessary protection. 3D scanning is performed on three positions frequently used by a catcher during the course of a game by putting markings on the subject's legs at 3cm intervals : a standing, a half squat with knees bent to 90 degrees and 120 degrees of knee flexion. Using data obtained from the 3D scan, we analyzed the changes in skin length, radii of curvatures, and cross-sectional shapes, depending on the degree of knee flexion. The results of the analysis were used to decide an on the ideal segmentation of the leg guards by modeling posture. Knee flexions to 90 degrees and to $120^{\circ}$ induced lengthwise extensions than a standing. In particular, the vertical length from the center of the leg increases to a substantially higher degree when compared to those increased from the inner and the outer side of the leg. The degree of extension is varied by positions. Therefore, the leg guards are segmented at points where the rate of increase changed. It resulted in a three-part segmentation of the leg guards at the thigh, the knee, and the shin. Since the 120 degree knee-flexion posture can accommodate other positions as well, the related 3D data are used for modeling Leg Guard (A) with the loft method. At the same time, Leg Guard (B) was modeled with two-part segmentation without separating the knee and the shin as in existing products. A biomechanical analysis of the new design is performed by simulating a 3D dynamic analysis. The analysis revealed that the three-part type (A) leg guards required less energy from the human body than the two-part type (B).