• Tunnel and Underground Space
• /
• v.33 no.6
• /
• pp.508-518
• /
• 2023

This paper presents a new approach for the automatic mapping of discontinuities in a tunnel face based on its 3D digital model reconstructed by LiDAR scan or photogrammetry techniques. The main idea revolves around the identification of discontinuity areas in the 3D digital model of a tunnel face by segmenting its 2D projected images using a deep-learning semantic segmentation model called U-Net. The proposed deep learning model integrates various features including the projected RGB image, depth map image, and local surface properties-based images i.e., normal vector and curvature images to effectively segment areas of discontinuity in the images. Subsequently, the segmentation results are projected back onto the 3D model using depth maps and projection matrices to obtain an accurate representation of the location and extent of discontinuities within the 3D space. The performance of the segmentation model is evaluated by comparing the segmented results with their corresponding ground truths, which demonstrates the high accuracy of segmentation results with the intersection-over-union metric of approximately 0.8. Despite still being limited in training data, this method exhibits promising potential to address the limitations of conventional approaches, which only rely on normal vectors and unsupervised machine learning algorithms for grouping points in the 3D model into distinct sets of discontinuities.

• Journal of Broadcast Engineering
• /
• v.28 no.3
• /
• pp.293-301
• /
• 2023

Unstandardized medical data collection and management are still being conducted manually, and studies are being conducted to classify CT data using deep learning to solve this problem. However, most studies are developing models based only on the axial plane, which is a basic CT slice. Because CT images depict only human structures unlike general images, reconstructing CT scans alone can provide richer physical features. This study seeks to find ways to achieve higher performance through various methods of converting CT scan to 2D as well as axial planes. The training used 1042 CT scans from five body parts and collected 179 test sets and 448 with external datasets for model evaluation. To develop a deep learning model, we used InceptionResNetV2 pre-trained with ImageNet as a backbone and re-trained the entire layer of the model. As a result of the experiment, the reconstruction data model achieved 99.33% in body part classification, 1.12% higher than the axial model, and the axial model was higher only in brain and neck in contrast classification. In conclusion, it was possible to achieve more accurate performance when learning with data that shows better anatomical features than when trained with axial slice alone.

• Journal of Korea Multimedia Society
• /
• v.6 no.7
• /
• pp.1286-1301
• /
• 2003

This paper introduces segmentation based environment modeling method and integration method using multiple environment map for constructing the realtime image-based panoramic navigation system. The segmentation-based environment modeling method is easy to implement on the environment map and can be used for environment modeling by extracting the depth value by the segmentation of the environment map. However, an environment model that is constructed using a single environment map has the problem of a blurring effect caused by the fixed resolution, and the stretching effect of the 3D model caused when information that does not exist on the environment map occurs due to the occlusion. In this paper, we suggest environment models integration method using multiple environment map to resolve the above problem. This method can express parallax effect and expand the environment model to express wide range of environment. The segmentation-based environment modeling method using multiple environment map can build a detail model with optimal resolution.

• Proceedings of the Korean Information Science Society Conference
• /
• 1998.10c
• /
• pp.591-593
• /
• 1998

실제감을 가지는 가상의 환경을 만들고, 사용자의 interaction에 맞추어 가상 환경과오브젝트를 조정하는 분야는 최근에 들어 많은 관심을 끌고 있다. 본 논문에서는 실제 영상을 이용하여 3차원 오브젝트 모델링 및 합성하는 방법에 대한 논의를 하였다. 실제 영상에서 깊이 정보를 얻어내는 방법으로는 다해상도 다기선 스테레오 정합 기법(multi-resolution multiple-baseline stereo matching technique : MR-MBS)을 이용하였으며, 실제 영상에서 생성된 모델을 Open Inventor 환경에서 모델을 재구성하고 performer를 이용하여, 최종적인 합성 영상을 만들어 내었다. 합성된 임의시점에서의 영상에서 사용자 조작에 따라 각각의 오브젝트를 조정 할 때 결과 영상을 실시간으로 얻을 수 있었고, 실제 영상의 텍스춰를 그대로 사용함으로써 실제감을 높일 수 있었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.10b
• /
• pp.619-621
• /
• 2003

CT, MRI와 같은 의료 영상을 3차원 재구성을 통해 3차원으로 가시화하는 기술은 의료 분야에서 진단과 진료에 많이 활용되고 있다. 이는 의사에게 정확한 진단과 시술에 대한 확신을 환자에게는 시술에 대한 이해와 신뢰를 심어준다. 치아의 경우 치과 진료가 개개의 치아에 이루어진다는 점을 고려하면 개개의 치아가 개별적으로 모델링 되어 3차원 상에 가시화 되어야 한다. Contour 기반 알고리즘은 2차원 단면 데이터로부터 고속 렌더링과 높은 품질의 3차원 모델 생성이 가능하다. 본 논문에서는 CT에서 추출한 치아 contour 데이터로부터 삼각형 패치를 생성하고, 어금니의 분기 문제를 해결하는 3차원 치아 모델 생성을 위한 contour 기반 알고리즘을 제안한다.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.7
• /
• pp.104-115
• /
• 1999

This paper proposes an efficient method for 3D modeling of a human face from trinocular images by reconstructing face surface using range data. By using a trinocular camera system, we mitigated the tradeoff between the occlusion problem and the range resolution limitation which is the critical limitation in binocular camera system. We also propose an MPC_MBS (Matching Pixel Count Multiple Baseline Stereo) area-based matching method to reduce boundary overreach phenomenon and to improve both of accuracy and precision in matching. In this method, the computing time can be reduced significantly by removing the redundancies. In the model generation sub-pixel accurate surface data are achieved by 2D interpolation of disparity values, and are sampled to make regular triangular meshes. The data size of the triangular mesh model can be controlled by merging the vertices that lie on the same plane within user defined error threshold.

• Journal of Korea Game Society
• /
• v.10 no.1
• /
• pp.147-156
• /
• 2010

In this paper, we propose 3D cloud animation by cloud modeling method of 2D images retrieved from a meteorological satellite. First, on the satellite images, we locate numerous control points to perform thin-plate spline warping analysis between consecutive frames for the modeling of cloud motion. In addition, the spectrum channels of visible and infrared wavelengths are used to determine the amount and altitude of clouds for 3D cloud image reconstruction. Pre-integrated volume rendering method is used to achieve seamless inter-laminar shades in real-time using small number of slices of the volume data. The proposed method could successfully construct continuously moving 3D clouds from 2D satellite images at an acceptable speed and image quality.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.34 no.5
• /
• pp.479-493
• /
• 2016

LiDAR(Light Detection and Ranging) data acquired from ALS(Airborne Laser Scanner) has been intensively utilized to reconstruct object models. Especially, researches for 3D modeling from LiDAR data have been performed to establish high quality spatial information such as precise 3D city models and true orthoimages efficiently. To reconstruct object models from irregularly distributed LiDAR point clouds, sensor calibration, noise removal, filtering to separate objects from ground surfaces are required as pre-processing. Classification and segmentation based on geometric homogeneity of the features, grouping and representation of the segmented surfaces, topological analysis of the surface patches for modeling, and accuracy assessment are accompanied by modeling procedure. While many modeling methods are based on the segmentation process, this paper proposed to extract key points directly for building modeling without segmentation. The method was applied to simulated and real data sets with various roof shapes. The results demonstrate feasibility of the proposed method through the accuracy analysis.

• Composites Research
• /
• v.34 no.1
• /
• pp.70-75
• /
• 2021

This paper presents a novel algorithm to reconstruct meso-scale representative volume elements (RVE), referring to experimentally observed features of Sheet Molding Compound (SMC) composites. Predicting anisotropic mechanical properties of SMC composites is challenging in the multiscale virtual test using finite element (FE) models. To this end, an SMC RVE modeler consisting of a series of image processing techniques, the novel reconstruction algorithm, and a FE mesh generator for the SMC composites are developed. First, micro-CT image processing is conducted to estimate probabilistic distributions of two critical features, such as fiber chip orientation and distribution that are highly related to mechanical performance. Second, a reconstruction algorithm for 3D fiber chip packing is developed in consideration of the overlapping effect between fiber chips. Third, the macro-scale behavior of the SMC is predicted by the multiscale analysis.

• Applied Microscopy
• /
• v.38 no.1
• /
• pp.51-61
• /
• 2008

This simulation study examined the effect of data acquisition error including the data type of TEM image, and incident beam intensity of the tilt series on 3D tomograms. Simulation was performed with the 3D head phantom model of Kak and Slaney, and the slightly modified 3D head phantom model with enhanced difference in absorption coefficients. Reconstructed tomogram for the original head phantom model using 8-bit gray-scale image was distorted with extremely high level of noise, while an acceptable result was obtained for the modified model. The results for the original model using wrong formulation for the transmitted beam intensity was proved to be incorrect. The high level of noise along the z direction was found in case of the modified model. On the other hand, the wrong value of incident beam intensity in both models gave distorted results. In order to reconstruct an artifacts-free 3D structure from the projections with invisible features in electron tomography, the 16-bit projection images should be used with the correct incident beam intensity which is applied to Beer's law.