http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
3D 딥러닝을 위한 CityJSON 기반의 점군 학습 데이터 생성
손경완,신영하,염재홍,이동천 한국측량학회 2022 한국측량학회지 Vol.40 No.6
DL (Deep Learning) has evolved and developed rapidly in various areas in the past decade since the concept ofANN (Artificial Neural Networks) was introduced in the 1940s. Recently, GeoAI (Geospatial Artificial Intelligence)is one of the newly emerging fields of geoinformatics to create geospatial products and to provide smart decisionmakingthrough the AI-based procedure including implementation of the DL models with model training. However,the major obstacle of the DL is to obtain and/or to create training data and labeling. In particular, labeling, is animportant task in deep learning that requires an enormous amount of time and effort, which in most cases is donemanually through visual judgement. The labeling process for the point cloud data is more challenging and timeconsumingthan for imagery. In this regard, the purpose of this paper is to present a method for systematically,efficiently, and automatically generating large amounts of training and label datasets required for 3D DL usingpoint cloud data such as LiDAR by utilizing CityJSON. In addition, surface normal vectors of the point cloudsthat could provide geometric information of the object were estimated. Utilizing the normal vectors along with the3D coordinates of the point clouds could improve DL model training performance. The datasets generated fromthe proposed method could be used for 3D DL model training to perform semantic, instance and part segmentationof buildings, and eventually 3D building modeling. In appendix, python codes for generating dataset based on theproposed method are provided. 딥러닝(DL: Deep Learning)은 1940년대에 인공신경망(ANN: Artificial Neural Network) 개념이 도입된 이후,특히 지난 10년 동안 다양한 분야에서 급속하게 발전과 진화를 계속하고 있다. 최근 GeoAI (Geospatial A rtificialIntelligence)는 AI 기반의 DL 기술을 공간정보에 구현하여 여러 성과물을 생성하고 지능적 의사결정을 제공하기 위해 새롭게 부상하고 있는 공간정보의 한 분야이다. 그러나 DL의 주요 장애물은 다량의 학습 데이터와 레이블을 획득하고 생성하는 것이다. 특히 레이블링은 많은 시간과 전문적 인력이 요구되는 중요한 과정으로 대부분의 경우 시각적판단에 의존하여 수동적인 방법으로 수행된다. 점군집 데이터의 레이블링은 영상에 비해 더 어렵고 많은 시간이 소요된다. 이러한 문제점을 해결하기 위해 본 논문은 CityJSON을 활용하여 LiDAR와 같은 점군 데이터를 이용하여 3차원 DL에 필요한 대용량의 학습 데이터와 레이블 셋을 체계적이고 효율적이며 자동으로 생성하는 방법을 제시하고있다. 또한 객체의 기하학적 정보를 제공할 수 있는 표면 법선벡터를 모든 점에서 추정하였다. 점군 데이터의 3차원 좌표와 법선벡터를 함께 학습에 사용하면, DL 모델의 학습성능을 향상시킬 수 있다. 제안한 방법에 의해 생성된 데이터셋을 점군 기반의 3차원 DL 모델 학습에 사용하면 건물의 의미적 분할과 객체 및 구성요소 분할을 수행하고 궁극적으로 3차원 건물 모델링을 수행할 수 있다. 제안한 방법을 수행하기 위한 주요 파이썬 코드를 부록에 제공하였다.
디지털 3차원 점군데이터 기반 암반 불연속면 거칠기 이방성 평가
김태현,김광염 한국암반공학회 2023 터널과지하공간 Vol.33 No.6
불연속면 거칠기는 암반의 기계적 특성에 큰 영향을 주며 열·수리 역학적 거동에도 많은 영향을 미치는요소이다. 본 연구에서는 입체사진측량기법을 이용하여 불연속면에 대한 3차원 점군 데이터를 생성시키고 이를 이용하여 불연속면의 거칠기 특성화를 수행하였다. 3차원 점군 데이터로 재생성된 불연속면프로파일과 프로파일 게이지를 이용하여 수동으로 측정한 프로파일을 비교하여 취득한 점군 데이터가암반면의 실제 형상을 정확하게 재현하였는지 평가하였다. 또한, 측정 프로파일수가 거칠기 평가에 미치는 영향에 대해 분석하였고, 거칠기의 이방성 평가방법을 제안하고 실제 암반 불연속면에 대한 거칠기이방성 평가를 수행하였다. The roughness of discontinuity significantly influences the mechanical characteristics of rock masses and extensively affects thermal and hydraulic behaviors. In this study, we utilized photogrammetry to generate 3D point cloud data for discontinuity and applied this data to characterize the roughness of discontinuity. The discontinuity profiles, reconstructed from the 3D point cloud data, were compared with those manually measured using a profile gauge. This comparison served to validate the accuracy and reliability of the acquired point cloud data in replicating the actual configurations of rock surfaces. Subsequent to this validation, influence of the number of profiles for representative JRC assessment was further investigated followed by suggestion of roughness anisotropy evaluation method with application of it to actual rock discontinuity surfaces.
3D 점군 데이터에서의 발 고유 변인 검출 시스템 설계
윤정록(Jeongrok Yun),김운용(Un Yong Kim),김회민(Hoemin Kim),전성국(Sungkuk Chun) 한국컴퓨터정보학회 2021 한국컴퓨터정보학회 학술발표논문집 Vol.29 No.2
최근 다양한 연령층과 직업군들 사이에서 기능성 신발에 대한 관심이 증대되고 있다. 그러나 기능성 신발 및 맞춤형 신발은 높은 가격대와 긴 제작 시간이 필요하다. 이러한 문제점은 3D 스캐너 도입으로 해결이 가능하나, 정확한 발 형상 측정이 가능한 3D 스캐너는 고가의 장비이기 때문에 매장별 보급이 어렵다. 본 논문은 기능성 신발의 보급을 위하여 저가형 3D 스캐너에서 정확한 발 고유 변인을 측정할 수 있는 시스템에 대해 서술한다. 이를 위해 이를 위해 저가형 Depth Camera를 이용한 저가형 3D 스캐너의 발 형상 3D 점군 데이터를 2차원으로 변형하고, 발 형태를 감싸는 최소 사각형(Min Area Rect)를 형성하여 발 안쪽점 및 발 가쪽점을 추정한다. 생성된 최소 사각형과 발 안쪽점 및 발 가쪽점 등은 발 고유 변인 측정의 기준이 된다. 실험 결과에서는 측정 기준을 이용하여 발 고유 변인인 발 길이, 발 너비, 발꿈치 너비, 발꿈치에서 발 안쪽점 및 발 가쪽점 길이 등 5가지 고유 변인을 측정하는 것을 보여준다.
점군 데이터를 활용한 옹벽의 단면 수치 정보 자동화 도출
한제희,장민서,한형서,조형준,신도형 한국BIM학회 2024 KIBIM Magazine Vol.14 No.2
The paper proposes a methodology that combines the Random Sample Consensus (RANSAC) algorithm and the Point Cloud Encoder-Decoder Network (PCEDNet) algorithm to automatically extract the length of infrastructure elements from point cloud data acquired through 3D LiDAR scans of retaining walls. This methodology is expected to significantly improve time and cost efficiency compared to traditional manual measurement techniques, which are crucial for the data-driven analysis required in the precisiondemanding construction sector. Additionally, the extracted positional and dimensional data can contribute to enhanced accuracy and reliability in Scan-to-BIM processes. The results of this study are anticipated to provide important insights that could accelerate the digital transformation of the construction industry. This paper provides empirical data on how the integration of digital technologies can enhance efficiency and accuracy in the construction industry, and offers directions for future research and application.
유동진(Dong-Jin Yoo) Korean Society for Precision Engineering 2006 한국정밀공학회지 Vol.23 No.9
In this paper a new approach which combines implicit surface scheme and NURBS surface interpolation method is proposed in order to generate a complete surface model from unorganized point cloud data. In the method a base surface was generated by creating smooth implicit surface from the input point cloud data through which the actual surface would pass. The implicit surface was defined by a combination of shape functions including quadratic polynomial function, cubic polynomial functions and radial basis function using adaptive domain decomposition method. In this paper voxel data which can be extracted easily from the base implicit surface were used in order to generate rectangular net with good quality using the normal projection and smoothing scheme. After generating the interior points and tangential vectors in each rectangular region considering the required accuracy, the NURBS surface were constructed by interpolating the rectangular array of points using boundary tangential vectors which assure C¹ continuity between rectangular patches. The validity and effectiveness of this new approach was demonstrated by performing numerical experiments for the various types of point cloud data.
시뮬레이션 데이터로 학습된 딥러닝 모델을 사용한 항공 라이다 데이터의 건물 모델링
신영하(Shin, Young-Ha),형성웅(Hyung, Sung-Woong),손경완(Son, Kyung-Wan),이동천(Lee, Dong-Cheon) 한국측량학회 2023 한국측량학회지 Vol.41 No.6
Photogrammetry is based on object modeling technology that extracts and measures important information about various objects and geographical features from images for 3D object reconstruction. Technological advances have made it possible to acquire various types of high quality data with inherent characteristics from different kinds of sensors. In particular, LiDAR (Light Detection and Ranging) which could directly obtain high-accuracy 3D information, is getting more widely used in various applications. This paper aims to develop a system that automatically models buildings in 3D using airborne LiDAR data. Roof surface segmentation, an important process of building modeling, was performed by deep learning. Deep learning model training requires a large amount of various data. To obtain enough amount of airborne LiDAR data for training deep learning model is difficult task. Therefore, simulated building dataset of point clouds was created for training the deep learning model. The simulated dataset consisting of 3D coordinates and normal vectors depicts buildings with various roof types. The roof surfaces were segmented by feeding airborne LiDAR data to the deep learning model trained with simulation dataset. Finally, building models with CityJSON format were generated by extracting model key points such as vertex and apex automatically from the segmented roof surfaces. In addition, a web system for processing and visualizing spatial information data was implemented to display results.
임의의 점 군 데이터로부터 쾌속조형을 위한 입력데이터의 자동생성
유동진(Dong-Jin Yoo) Korean Society for Precision Engineering 2007 한국정밀공학회지 Vol.24 No.11
In order to generate the input data for rapid prototyping, a new approach which is based on the implicit surface interpolation method is presented. In the method a surface is reconstructed by creating smooth implicit surface from unorganized cloud of points through which the surface should pass. In the method an implicit surface is defined by the adaptive local shape functions including quadratic polynomial function, cubic polynomial function and RBF(Radial Basis Function). By the reconstruction of a surface, various types of error in raw STL file including degenerated triangles, undesirable holes with complex shapes and overlaps between triangles can be eliminated automatically. In order to get the slicing data for rapid prototyping an efficient intersection algorithm between implicit surface and plane is developed. For the direct usage for rapid prototyping, a robust transformation algorithm for the generation of complete STL data of solid type is also suggested.