• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.01a
• /
• pp.369-372
• /
• 2023

본 논문에서는 삼각형 구조로 구성된 3차원 메쉬(Mesh)에서 합성곱 신경망(Convolution Neural Network, CNN)을 응용하여 정확도가 높은 새로운 학습 표현 기법을 제시한다. 우리는 메쉬를 구성하고 있는 폴리곤의 edge와 face의 로컬 특징을 기반으로 학습을 진행한다. 일반적으로 딥러닝은 인공신경망을 수많은 계층 형태로 연결한 기법을 말하며, 주요 처리 대상은 1, 2차원 데이터 형태인 오디오 파일과 이미지였다. 인공지능에 대한 연구가 지속되면서 3차원 딥러닝이 도입되었지만, 기존의 학습과는 달리 3차원 딥러닝은 데이터의 확보가 쉽지 않다. 혼합현실과 메타버스 시장의 확대로 인해 3차원 모델링 시장이 증가하고, 기술의 발전으로 데이터를 획득할 수 있는 방법이 생겼지만, 3차원 데이터를 직접적으로 학습에 이용하는 방식으로 적용하는 것은 쉽지 않다. 그렇게 때문에 본 논문에서는 산업 현장에서 이용되는 데이터인 메쉬 구조를 폴리곤의 최소 단위인 삼각형 형태로 구성하여 학습 데이터를 구성해 기존의 방법보다 정확도가 높은 학습 기법을 제안한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2018.11a
• /
• pp.53-54
• /
• 2018

주어진 영상으로부터의 3 차원 얼굴 모델링은 얼굴 분석, 애니메이션, 생체 인식 등의 많은 컴퓨터비전 및 그래픽스 응용분야에서 중요한 역할을 하고 있다. 그 중에서도 헤어 영역은 얼굴에 비해 모양의 다양성과 모델의 복잡도가 현저히 높다. 기존의 연구는 주로 얼굴 영역에 한정한 3 차원 얼굴 모델링을 중심으로 이루어졌지만 헤어 모델링은 중요하게 다루지 않고 있는 경우가 많다. 본 논문에서는 심층인공신경망의 일종인 FCN (fully connected network)을 이용하여 인물 영상에서 헤어 부분을 영역화하고 그와 가장 유사한 3D 헤어 모델을 데이터베이스에서 검색하여 3 차원 얼굴 모델에 증강함으로써 완전한 얼굴 모델링을 수행하는 방법을 제안한다. 이는 FCN 을 이용하여 다양한 인물 영상에 대하여 네트워크 학습을 수행하는 과정과 3D 헤어 데이터베이스의 구축 과정을 포함한다. 실험 결과 적절한 수준의 헤어 모델이 3 차원 얼굴 모델링 결과물에 증강됨을 확인하였다.

• Advances in Computational Design
• /
• v.5 no.3
• /
• pp.277-289
• /
• 2020

This article describes the technology for constructing of a multiply-connected three-dimensional area's finite element representation. Representation of finite-element configuration of an area is described by a discrete set that consist of the number of nodes and elements of the finite-element grid, that are orderly set of nodes' coordinates and numbers of finite elements. Corresponding theorems are given, to prove the correctness of the solution method. The adequacy of multiply-connected area topology's finite element model is shown. The merging of subareas is based on the criterion of boundary nodes' coincidence by establishing a simple hierarchy of volumes, surfaces, lines and points. Renumbering nodes is carried out by the frontal method, where nodes located on the outer edges of the structure are used as the initial front.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.42 no.6
• /
• pp.75-82
• /
• 2005

The robust feature vector selection method for the multiple frontal face detection is proposed in this paper. The proposed feature vector for the training and classification are integrated by means, amplitude projections, and its 1D Harr wavelet of the input image. And the statistical modeling is performed both for face and nonface classes. Finally, the estimated probability density functions (PDFs) are applied for the detection of multiple frontal faces in the still image. The proposed method can handle multiple faces, partially occluded faces, and slightly posed-angle faces. And also the proposed method is very effective for low quality face images. Experimental results show that detection rate of the propose method is $98.3\%$ with three false detections on the testing data, SET3 which have 227 faces in 80 images.

• Korean Journal of Computational Design and Engineering
• /
• v.18 no.2
• /
• pp.93-103
• /
• 2013

In the situation where there should be a gap between the design environments of 2D drawing and 3D BIM modeling, it is unavoidable aspect to face confusion caused by parallel usage of both technologies unless the project delivery systems are entirely upgraded to 3D. Therefore, it is critical to make efforts to minimize this confusion. Meanwhile, detailed drawings have very important roles in the construction process because they do not only concretely express construction methods for a given design plan but also provide basis for quantity takeoff and construction cost estimation etc. This paper suggests a plan to standardize detailed information based on BIM standard framework in order to make BIM realize the information. In the short term, this plan provides improved environment where both BIM and CAD technologies can be applied and, in the long term, it provides bases for BIM to replace CAD technology.

• Journal of KIISE:Computer Systems and Theory
• /
• v.32 no.10
• /
• pp.520-529
• /
• 2005

A feature-based 3D modeling algorithm is presented in this paper. Since conventional methods use depth-based techniques, they need much time for the image matching to extract depth information. Even feature-based methods have less computation load than that of depth-based ones, the calculation of modeling error about whole pixels within a triangle is needed in feature-based algorithms. It also increase the computation time. Therefore, the proposed algorithm consists of three phases, which are an initial 3D model generation, model evaluation, and model refinement phases, in order to acquire an efficient 3D model. Intensity gradients and incremental Delaunay triangulation are used in the Initial model generation. In this phase, a morphological edge operator is adopted for a fast edge filtering, and the incremental Delaunay triangulation is modified to decrease the computation time by avoiding the calculation errors of whole pixels and selecting a vertex at the near of the centroid within the previous triangle. After the model generation, sparse vertices are matched, then the faces are evaluated with the size, approximation error, and disparity fluctuation of the face in evaluation stage. Thereafter, the faces which have a large error are selectively refined into smaller faces. Experimental results showed that the proposed algorithm could acquire an adaptive model with less modeling errors for both smooth and abrupt areas and could remarkably reduce the model acquisition time.

• Journal of Information Processing Systems
• /
• v.17 no.4
• /
• 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.

• IE interfaces
• /
• v.12 no.2
• /
• pp.205-209
• /
• 1999

In reverse engineering area, the reconstruction of surfaces from scanned or digitized data is being developed, but geometric model of existing objects is not available in industries. This paper presents the new approach to the reconstruction of surface technique. A proposed methodology finds base geometry and blends surface between them. Each based geometry is divided by tri-angular patches which are compared with their normal vector for face grouping. Each group is categorized analytical surface such as a part of cylinder, sphere and cone, and plane shapes to represent the based geometry surface. And then, each based geometry surface is implemented to the infinitive surface. Infinitive surface's intersections are trimmed by boundary representation model reconstruction. This method has several benefits such as time efficiency and automatic functional modeling system in reverse engineering. Especially, it can be directly applied 3D fax and 3D copier.

• Journal of KIISE:Computing Practices and Letters
• /
• v.5 no.4
• /
• pp.385-394
• /
• 1999

제품을 설계하는 디자이너나 엔지니어는 많은 시간과 노력을 들이지 않고서 그들이 설계한 3차원 제품 모델에 대한 사실적인 이미지를 원한다. 디자인 프로세스의 초기인 개념 설계에서부터 설계검증, 그리고 가공 과정에서 사실적인 이미지가 매우 유용하므로, 대부분의 주요 CAD 제작사는 그들의 CAD 소프트웨어에 고급 렌더링 기능을 추가하고 있다. 상용의 CAD/CAM 모델러에서는 NURB 곡면을 기초로 모델링을 수행하므로, NURB 곡면을 렌더링할 수 있는 패키지가 필요하다. VIF(Visual InterFace) 렌더링 라이브러리는 A-buffer 방식과 Ray tracing 방식의 두 가지 고급 렌더링 모드를 제공한다. 다각형은 물론 NURB 곡면을 입력으로 받아 사용자가 설정한 표면의 각종 계수, 원하는 view와 설정된 광원에 따라 이미지를 만들고 다양한 형태로 출력시킬 수 있는 다양한 기능을 제공한다. 본 논문에서는 VIF 렌더링 라이브러리에 대한 구조와 기능별로 분류된 함수에 대하여 설명하며, 실제로 CAD/CAM 시스템과 통합되어 구상설계에서부터 3차원 설계 모델링에 이르기까지의 제조공정에서 설계검증 툴로써 어떻게 활용되고 있는가에 대하여 기술한다.Abstract Engineers and industrial designers want to produce a realistic-looking images of a 3D model without spending a lot of time and money. Photo-realistic images are so useful from the conceptual design, through its verification, to the machining, that most major CAD venders offer built-in as well as add-on photo-realistic rendering capability to their core CAD software. Since 3D model is consists of a set of NURB surfaces in commercial CAD packages, we need a renderer which handles NURB surface as well as other primitives.A new rendering library called VIF (Visual InterFace) provides two photo-realistic rendering modes: A-buffer and Ray tracing. As an input data it takes NURB surfaces as well as polygonal data and produces images in accordance with the surface parameters, view and lights set by user and outputs image with different formats. This paper describes the overall architecture of VIF and its library functions classified by their functionalities, and discusses how VIF is used as a graphical verification tool in manufacturing processes from the conceptual design to 3D modeling.

• Journal of the Korea Society for Simulation
• /
• v.32 no.3
• /
• pp.1-8
• /
• 2023

Modeling of artificial targets in Synthetic Aperture radar (SAR) mainly simulates radar signals reflected from the faces and edges of the 3D Computer Aided Design (CAD) model with a ray-tracing method, and modeling of the clutter on the Earth's surface uses a method of distinguishing types with similar distribution characteristics through statistical analysis of the SAR image itself. In this paper, man-made targets on the surface and background clutter on the terrain are integrated and made into a three-dimensional (3D) point cloud scatterer model, and SAR image were created through computational signal processing. The results of the SAR Stripmap image generation of the actual automobile based SAR radar system and the results analyzed using EM modeling or statistical distribution models are compared with this 3D point cloud scatterer model. The modeling target is selected as an bridge because it has the characteristic of having both water surface and ground terrain around the bridge and is also a target of great interest in both military and civilian use.