• 전자공학회논문지CI
• /
• 제47권3호
• /
• pp.29-35
• /
• 2010

본 논문에서는 한 장의 정지 영상에서 학습을 통한 방법으로 깊이 정보를 추정하는데 사용되어지는 특징 정보를 PCA(Principal Component Analaysis)기반으로 축소하여 깊이 정보의 정확성을 향상시키는 방법에 대하여 기술한다. 정지 영상에서 깊이 정보를 추정하기 위하여 이미지의 에너지 값과 기울기와 같은 특징을 추출하며 특징들의 관계를 이용하여 각 영역의 깊이 정보를 추정한다. 이 때 영상 필터를 사용하여 많은 특징을 추출하지만 특징의 중요성을 판단하지 않고 모두 사용하면 오히려 성능에 좋지 않은 영향을 미친다. 본 논문에서는 한 장의 정지 영상의 깊이 추정을 위해 PCA를 기반으로 중요도를 판단하여 특징 벡터의 차원을 줄이고 깊이를 정확하게 추정할 수 있는 방법에 대하여 제안한다. 제안한 방법을 스탠포드 대학의 평가 데이터로 실험한 결과, 깊이를 추정하는데 있어서 전체 특징 벡터의 30%만을 이용하여 평균 0.4%에서 최대 2.5%의 정확도가 향상되었다.

• 한국멀티미디어학회논문지
• /
• 제19권9호
• /
• pp.1647-1658
• /
• 2016

RGB+D database has been widely used in object recognition, object tracking, robot control, to name a few. While rapid advance of active depth sensing technologies allows for the widespread of indoor RGB+D databases, there are only few outdoor RGB+D databases largely due to an inherent limitation of active depth cameras. In this paper, we propose a novel method used to build outdoor RGB+D databases. Instead of using active depth cameras such as Kinect or LIDAR, we acquire a pair of stereo image using high-resolution stereo camera and then obtain a depth map by applying stereo matching algorithm. To deal with estimation errors that inevitably exist in the depth map obtained from stereo matching methods, we develop an approach that estimates confidence of depth maps based on unsupervised learning. Unlike existing confidence estimation approaches, we explicitly consider a spatial correlation that may exist in the confidence map. Specifically, we focus on refining confidence feature with the assumption that the confidence feature and resultant confidence map are smoothly-varying in spatial domain and are highly correlated to each other. Experimental result shows that the proposed method outperforms existing confidence measure based approaches in various benchmark dataset.

• 제어로봇시스템학회논문지
• /
• 제19권5호
• /
• pp.457-461
• /
• 2013

This paper proposes an image feature-based real-time RGB-D (Red-Green-Blue Depth) 3D SLAM (Simultaneous Localization and Mapping) system. RGB-D data from Kinect style sensors contain a 2D image and per-pixel depth information. 6-DOF (Degree-of-Freedom) visual odometry is obtained through the 3D-RANSAC (RANdom SAmple Consensus) algorithm with 2D image features and depth data. For speed up extraction of features, parallel computation is performed with GPU acceleration. After a feature manager detects a loop closure, a graph-based SLAM algorithm optimizes trajectory of the sensor and builds a 3D point cloud based map.

• ETRI Journal
• /
• 제44권1호
• /
• pp.38-50
• /
• 2022

This paper presents a study on the use of encoder-derived features in decoder-side depth estimation. The scheme of multiview video encoding does not require the transmission of depth maps (which carry the geometry of a three-dimensional scene) as only a set of input views and their parameters are compressed and packed into the bitstream, with a set of features that could make it easier to estimate geometry in the decoder. The paper proposes novel recursive block splitting for the feature extraction process and evaluates different scenarios of feature-driven decoder-side depth estimation, performed by assessing their influence on the bitrate of metadata, quality of the reconstructed video, and time of depth estimation. As efficient encoding of multiview sequences became one of the main scopes of the video encoding community, the experimental results are based on the "geometry absent" profile from the incoming MPEG Immersive video standard. The results show that the quality of synthesized views using the proposed recursive block splitting outperforms that of the state-of-the-art approach.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2022년도 추계학술대회
• /
• pp.256-257
• /
• 2022

본 논문에서는 단일 모바일 디바이스의 움직임을 통해 3차원 특징점을 추출하는 방법에 대해 소개한다. 단안 카메라를 이용해 카메라 움직임에 따라 2D 영상을 획득하고 Baseline을 추정한다. 특징점 기반의 스테레오 매칭을 진행한다. 특징점과 디스크립터를 획득하고 특징점을 매칭한다. 매칭된 특징점을 이용해 디스패리티를 계산하고 깊이값을 생성한다. 3차원 특징점은 카메라 움직임에 따라 업데이트 된다. 마지막으로 장면 전환 검출을 이용하여 장면 전환시 특징점을 리셋한다. 위 과정을 통해 특징점 데이터베이스에 평균 73.5%의 저장공간 추가 확보를 할 수 있다. TUM Dataset의 Depth Ground truth 값과 RGB 영상으로 제안한 알고리즘을 적용하여 3차원 특징점 결과와 비교하여 평균 26.88mm의 거리 차이가 나는것을 확인하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제10권6호
• /
• pp.2730-2747
• /
• 2016

Due to the limitation of the bandwidth resource and capture resolution of depth cameras, low resolution depth maps should be up-sampled to high resolution so that they can correspond to their texture images. In this paper, a novel depth map up-sampling algorithm is proposed by exploiting the fractal internal self-referential feature. Fractal parameters which are extracted from a depth map, describe the internal self-referential feature of the depth map, do not introduce inherent scale and just retain the relational information of the depth map, i.e., fractal transforms provide a resolution-independent description for depth maps and could up-sample depth maps to an arbitrary high resolution. Then, an enhancement method is also proposed to further improve the performance of the up-sampled depth map. The experimental results demonstrate that better quality of synthesized views is achieved both on objective and subjective performance. Most important of all, arbitrary resolution depth maps can be obtained with the aid of the proposed scheme.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권5호
• /
• pp.2327-2347
• /
• 2018

Human actions can be recognized from depth sequences. In the proposed algorithm, we initially construct depth, motion maps (DMM) by projecting each depth frame onto three orthogonal Cartesian planes and add the motion energy for each view. The body part of the action (BPoA) is calculated by using bounding box with an optimal window size based on maximum spatial and temporal changes for each DMM. Furthermore, feature vector is constructed by using BPoA for each human action view. In this paper, we employed an ensemble based learning approach called Rotation Forest to recognize different actions Experimental results show that proposed method has significantly outperforms the state-of-the-art methods on Microsoft Research (MSR) Action 3D and MSR DailyActivity3D dataset.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1997년도 한국자동제어학술회의논문집; 한국전력공사 서울연수원; 17-18 Oct. 1997
• /
• pp.1848-1851
• /
• 1997

This paper presents the development of depth detection algorithm for the 3D Endoscopic Data using a stereo matchod and depth calcuation.

• 수산해양기술연구
• /
• 제52권2호
• /
• pp.103-110
• /
• 2016

This study aims to estimate the species, size and shape of fish using a non-contact 3 dimensional pattern laser so that this preliminary test was carried out to understand the structural feature and length of goldfish according to water turbidity and depth in the aquacultural tank. 3-D pattern laser could clearly detect its morphological shape except the caudal fin due to soft tissue. Since the sensing strength of line laser light according to depth has sufficient power, it is possible to measure its depth and structural feature in the detected range. The result showed that the measured error of individual's fork length was less than ${\pm}1%$ in the water using 3-D pattern laser, when compared with the measured value in the air.

• 한국멀티미디어학회논문지
• /
• 제24권8호
• /
• pp.1026-1034
• /
• 2021

Tire wear and defect are important factors for safe driving condition. These defects are generally inspected by some specialized experts or very expensive equipments such as stereo depth camera and depth gauge. In this paper, we propose tire safety vision inspector based on deep neural network (DNN). The status of tire wear is categorized into three: 'safety', 'warning', and 'danger' based on depth of tire tread. We propose an attention mechanism for emphasizing the feature of tread area. The attention-based feature is concatenated to output feature maps of the last convolution layer of ResNet-101 to extract more robust feature. Through experiments, the proposed tire wear classification model improves 1.8% of accuracy compared to the existing ResNet-101 model. For detecting the tire defections, the developed tire defect detection model shows up-to 91% of accuracy using the Mask R-CNN model. From these results, we can see that the suggested models are useful for checking on the safety condition of working tire in real environment.