Earticle

댐은 생활에 필요한 수자원을 안정적으로 공급하고, 집중호우에 의한 피해를 저감하기 위해 건설된다. 댐의 건설은 많은 토지, 가옥 등이 수몰되고 주변지역의 환경에도 큰 영향을 주기 때문에 최근에는 댐 주변지역에 대한 공간정보의 분석 및 활용이 요구되고 있다. 이에 본 연구에서는 무인항공기를 활용하여 댐건설로 인한 수몰지역에 대한 공간정보를 구축하고 효과적인 공간정보 분석 방안을 제시하고자 하였다. 연구를 통해 댐 및 주변지역에 대한 정사영상과 DSM의 공간정보를 효과적으로 구축할 수 있었다. 또한 DSM에 대한 분석을 통해 댐의 담수에 따른 침 수 지역의 변화와 수량을 효과적으로 산정할 수 있었다. 무인항공시스템을 활용하여 공간정보를 구축하고, 이에 대한 분석을 기반으로 효과적인 댐 주변지역의 관리 방안 제시할 수 있었으며, 연구에서 수행된 공간정보 구축 및 분석 방법은 관련 분야의 기초자료로 활용이 기대된다.

Dam is built to supply water necessary for life stably and reduce the damage caused by heavy rains. Recently there has been required the analysis and utilization of spatial information on the area around the dam because construction of the dam gives a great impact on the environment of the surrounding area. In this study, we build spatial information about the submerged district due to dam construction using drone and propose the effective method for analysis of the spatial information. As a result, orthoimage and DSM of study area were constructed effectively. Change of submerged district of this area was calculated according to the analysis of spatial information. Building and analyzing spatial information carried out in this study are expected to be utilized as the basis in related fields.

최근 UAM(Urban Air Mobility)에 대한 관심이 도시의 교통 혼잡과 대기오염 문제 해결 방안으로 급증하고 있다. 하지만 UAM의 효율적인 운영을 위해서는 3D Point Cloud 데이터의 정확한 처리가 필요하며, 특히 지면과 객체를 분리하는 문제가 중요하다. 본 논문은 UAM 환경의 동적이고 복잡한 특성을 고려하여 지면과 객체를 효과적으로 분리하는 방법을 제안하고 검증한다. 우리의 접근 방식은 MEMS 센서로부터 얻은 자세 정보와 RANSAC을 이용한 지면 평면 추정을 결합하여, GPS 오차에 크게 영향 받지 않는 지면/객체 분리를 가능하게 한다. 시뮬레이션 결과는 이 방법이 UAM 환경에서 효과적으로 작동함을 보여주며, 도심 항공 모빌리티의 안전성과 효율성을 향상시키는 중요한 단계를 제시한다. 향후 연구는 이 알고리즘의 정확성을 높이고 다양한 UAM 환경에서 성능을 평가하며, 실제 드론 테스트를 진행할 예정이다.

Recently, interest in UAM (Urban Air Mobility) has surged as a critical solution to urban traffic congestion and air pollution issues. However, efficient UAM operation requires accurate 3D Point Cloud data processing, particularly in separating the ground and objects. This paper proposes and validates a method for effectively separating ground and objects in a UAM environment, taking into account its dynamic and complex characteristics. Our approach combines attitude information from MEMS sensors with ground plane estimation using RANSAC, allowing for ground/object separation that isless affected by GPS errors. Simulation results demonstrate that this method effectively operates in UAM settings, marking a significant step toward enhancing safety and efficiency in urban air mobility. Future research will focus on improving the accuracy of this algorithm, evaluating its performance in various UAM scenarios, and proceeding with actual drone tests.

In this study, a complete 3D surface reconstruction method is proposed based on the concept that the vertices, of surface model can be completely matched to the unstructured point cloud. In order to generate the initial mesh model from the point cloud, the mesh subdivision of bounding box and shrink-wrapping algorithm are introduced. The control mesh model for well representing the topology of point cloud is derived from the initial mesh model by using the mesh simplification technique based on the original QEM algorithm, and the parametric surface model for approximately representing the geometry of point cloud is derived by applying the local subdivision surface fitting scheme on the control mesh model. And, to reconstruct the complete matching surface model, the insertion of isolated points on the parametric surface model and the mesh optimization are carried out. Especially, the fast 3D surface reconstruction is realized by introducing the voxel-based nearest-point search algorithm, and the simulation results reveal the availability of the proposed surface reconstruction method.

In the rapidly evolving metaverse ecosystem, video streaming plays a vital role, offering immersive, on-demand experiences and engaging users in dynamic 3D virtual environments. Point cloud video streaming enables detailed 3D representations for applications like 3D scanning, AR, and VR. However, high data demands and complex encoding pose challenges, requiring efficient resource and latency management. As a result, user experience (UX) is critical in the metaverse, where metrics like real-time feedback, Quality of Experience (QoE), and adaptive UX design are essential for sustaining engagement in immersive 3D environments. Therefore, this paper closely examines the latest developments in point cloud video streaming, specifically focusing on user-centric, AI-driven, and low-latency techniques, transcending traditional video streaming methods. We investigate a variety of techniques, pinpoint obstacles, and chart prospective routes for elevating both the quality and depth of immersive experiences. Furthermore, our work includes a detailed qualitative comparison of these streaming paradigms. We also shed light on pivotal metaverse enhancers, such as Mobile Edge Computing (MEC), Visible Light Communication (VLC), advanced network coding techniques, and the integration of the Data Plane Development Kit (DPDK). These enablers?particularly when combined with DPDK’s capabilities in efficient data packet processing?have the potential to optimize resource allocation, strengthen reliability, reduce latency, and improve the sustainability of point cloud video streaming. Our study, therefore, offers a holistic overview of the advancing area of point cloud video streaming and outlines paths for future research and development.

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.

In this study, we proposed a new approach to segment ground and nonground points gained from a 3D laser range sensor. The primary aim of this research was to provide a fast and effective method for ground segmentation. In each frame, we divide the point cloud into small groups. All threshold points and start-ground points in each group are then analyzed. To determine threshold points we depend on three features: gradient, lost threshold points, and abnormalities in the distance between the sensor and a particular threshold point. After a threshold point is determined, a start-ground point is then identified by considering the height difference between two consecutive points. All points from a start-ground point to the next threshold point are ground points. Other points are nonground. This process is then repeated until all points are labelled.

In point cloud registration, the combination of global and local features provides a comprehensive representation of point cloud data. Previous registration methods predominantly relied on a single type of feature. For instance, correspondence matching-based methods require stringent criteria for the uniqueness of input point clouds, and the quality of feature extraction significantly influences the accuracy of registration results. Global feature-based methods capture the overall geometric information of the inputs; however, the absence of local information often leads to the neglect of the adverse effects caused by non-overlapping points. In this paper, we present DFPINet, an iterative network based on dual feature extraction and progressive feature interaction for partial-to-partial point cloud registration of small-scale object point clouds. We use a dual branch structure to individually extract both global and local features, fully utilizing the geometric characteristics of the point cloud. Additionally, the progressive feature interaction operation enhances feature connections and mitigates the effect of non-overlapping points. Experimental results demonstrate that our method surpasses existing registration approaches in both accuracy and robustness.

The premise of this research is the recent advancement of Building Information Modeling(BIM) Technology and Laser Scanning Technology(3D Scanning). The purpose of the paper is to amplify the potential offered by the combination of BIM and Point Cloud Data (PCD) for structural analysis. Today, enormous amounts of construction site data can be potentially categorized and quantified through BIM software. One of the extraordinary strengths of BIM software comes from its collaborative feature, which can combine different sources of data and knowledge. There are vastly different ways to obtain multiple construction site data, and 3D scanning is one of the effective ways to collect close-to-reality construction site data. The objective of this paper is to emphasize the prospects of pre-scanning and post-scanning automation algorithms. The research aims to stimulate the recent development of 3D scanning and BIM technology to develop Scan-to-BIM. The paper will review the current issues of Scan-to-BIM tasks to achieve As-Built BIM and suggest how it can be improved. This paper will propose a method of coordinating and utilizing PCD for construction and structural analysis during construction.

Cloud computing is an attractive solution that can provide low cost storage and powerful processing capabilities for government agencies or enterprises of small and medium size. Yet the confidentiality of information should be considered by any organization migrating to cloud, which makes the research on relational database system based on encryption schemes to preserve the integrity and confidentiality of data in cloud be an interesting subject. So far there have been various solutions for realizing SQL queries on encrypted data in cloud without decryption in advance, where generally homomorphic encryption algorithm is applied to support queries with aggregate functions or numerical computation. But the existing homomorphic encryption algorithms cannot encrypt floating-point numbers. So in this paper, we present a mechanism to enable the trusted party to encrypt the floating-points by homomorphic encryption algorithm and partial trusty server to perform summation on their ciphertexts without revealing the data itself. In the first step, we encode floating-point numbers to hide the decimal points and the positive or negative signs. Then, the codes of floating-point numbers are encrypted by homomorphic encryption algorithm and stored as sequences in cloud. Finally, we use the data structure of DoubleListTree to implement the aggregate function of SUM and later do some extra processes to accomplish the summation.

LiDAR는 자율주행차량이 주변의 객체를 정확하게 식별하고 안전을 강화하여 운전자를 지원할 수 있는 중요한 기술로 많은 ADAS(Advanced Driver Assistance System) 및 자율주행 응용 프로그램에서 차량의 안전성과 정확성을 향상시키는 도구로 사 용되고 있다. LiDAR를 통해 생성된 Point Cloud Data를 처리하기 위한 다양한 방법론들이 연구되고 있으며 특히, 딥러닝 분야 에서도 활발하게 연구가 진행되고 있다. 본 연구에서는 이러한 LiDAR의 Point Cloud Data(PCD)를 처리하기 위해서 PCD의 지역적인 특징을 수학적으로 계산하고 직 접 기계에 입력할 수 있는 기법인 Topological Data Analysis(TDA)를 사용하고자 하며 이를 통해 기존의 PCD를 다루는 딥러닝 모델인 PointNet, DGCNN 등에서 발생했던 문제인 로컬 구조와 패턴에 대한 인식, 객체의 스케일과 회전 등의 변형에 대한 민 감성을 해결해보고자 한다. 이를 통해 보다 정교한 인공지능 모델을 구축하고 실제 수집된 데이터에 대해 강건하게 대응할 수 있을 것으로 기대된다.

Precise forest inventory is crucial for sustainable forest management. Recently, LiDAR technology has been widely applied to forest inventory as it provides highly accurate measurements of tree height, canopy structure, and terrain, enabling efficient data collection and analysis. This study aims to (i) determine point cloud density of ground and vegetation layers classified from Handheld Mobile Laser Scanning (HMLS), Airborne Laser Scanning (ALS), and Integrated ALS and HMLS; (ii) compare DBH and tree height derived from HMLS, ALS, and Integrated ALS and HMLS, and (iii) analyse applicability of integrating HMLS and ALS scanning methods in estimate individual tree attributes of pine forests in South Korea. As a forthcoming LiDAR application, the fusion of ALS and HMLS method could transform forest inventory methods, overcoming the challenges of traditional ground-based surveys and enabling faster and more precise forest evaluations, which contributes to forest ecosystem assessment and carbon stock estimation.

MPEG-I (Immersive) 표준의 하나로 포인트 클라우드 압 축 (Point Cloud Coding/Compression) 기술의 표준화가 시작단계에 있다. 비디오와 그래픽스 부분의 접점이 AR/VR 부문에서 확대되고 있는 가운데, 공간상의 화소로 볼 수 있는 점 구름 데이터의 압축 기술에 대한 표준화에 대한 요구 또한 과거 그 어느 때보다 높다. 본 고에서는 점 구름 압축 표준화 에 대한 현재의 동향과 향후 전개 방향에 대하여 소개하고자 한다

Three-dimensional (3D) point clouds provide detailed geometric understanding of real-world environments but remain challenging to process due to their sparse and unordered nature. Contrastive learning has emerged as a powerful self-supervised approach for learning representations from unlabeled 3D point cloud data. At the core of these methods lie encoder architectures that project raw points into discriminative latent spaces. This brief survey highlights major encoder families used in 3D contrastive learning and analyzes their design principles, strengths, and limitations. We further discuss how encoder choice influences downstream performance and outline research trends toward efficient, multimodal, and real-time contrastive frameworks.

This review study discusses the utilization of Terrestrial Laser Scanner (TLS) 3D point cloud data in representing geometry for structural health monitoring (SHM) applications, particularly the so-called Scan-to-FEM method. The existing studies related to the topic are intensively presented and discussed along with the advantages and disadvantages of the modelling methods: meshing by whole and cross-section. Two case studies are conducted to show both modelling methods used for geometry representation for Scan-to-FEM method. Meshing by whole and cross-section methods result in the relative errors of respective 2.79% and 1.4% in the maximum displacement vector sum, compared to their reference models constructed by ABAQUS. It is also seen that the density and clarity of 3D point cloud data can bring well-shaped geometry for FE analysis and consequently produce good accuracy.

인공지능을 이용하여 예측이나 분석하는 기술은 지속적으로 발전하고 있지만, 의사결정 과정을 명확히 해석하지 못하는 블랙박스 문제가 존재한다. 따라서 인공지능 모델의 의사결정 과정에서 사용자의 입장에서 해석이 불가능하여 결과를 신뢰할 수 없는 문제가 발생한다. 본 연구에서는 인공지능의 문제점과 이를 해결하기 위한 블록체인을 활용한 설명 가능한 인공지능에 대해 연구를 진행하였다. 블록체인을 이용해서 설명 가능한 인공지능 모델의 의사결정 과정에서 의 데이터를 타임스탬프 등을 이용하여 부분별로 블록체인에 저장한다. 블록체인을 이용하여 저장된 데이터의 위변조 방지를 제공하고 블록체인의 특성상 사용자는 블록에 저장된 의사결정 과정등의 데이터를 자유롭게 접근할 수 있다. 설명 가능한 인공지능 모델의 구축이 힘든 것은 기존 모델의 복잡성이 큰 부분을 차지한다. 따라서 포인트 클라우드를 활용해 서 3차원 데이터 처리와 가공과정의 효율성을 높여서 의사결정 과정을 단축해 설명 가능한 인공지능 모델의 구축을 원활 하게 한다. 블록체인에 데이터 저장과정에서 데이터 위변조가 발생할 수 있는 오라클 문제를 해결하기 위해 저장과정에 중간자를 거치는 블록체인 기반의 설명 가능한 인공지능 모델을 제안하여 인공지능의 블랙박스 문제를 해결하였다.

Although the technology for prediction or analysis using artificial intelligence is constantly developing, a black-box problem does not interpret the decision-making process. Therefore, the decision process of the AI model can not be interpreted from the user's point of view, which leads to unreliable results. We investigated the problems of artificial intelligence and explainable artificial intelligence using Blockchain to solve them. Data from the decision-making process of artificial intelligence models, which can be explained with Blockchain, are stored in Blockchain with time stamps, among other things. Blockchain provides anti-counterfeiting of the stored data, and due to the nature of Blockchain, it allows free access to data such as decision processes stored in blocks. The difficulty of creating explainable artificial intelligence models is a large part of the complexity of existing models. Therefore, using the point cloud to increase the efficiency of 3D data processing and the processing procedures will shorten the decision-making process to facilitate an explainable artificial intelligence model. To solve the oracle problem, which may lead to data falsification or corruption when storing data in the Blockchain, a blockchain artificial intelligence problem was solved by proposing a blockchain-based explainable artificial intelligence model that passes through an intermediary in the storage process.

Presented in the paper is an algorithm to generate a section curve by slicing a point cloud including tens of thousands of points. Although, there have been previous research results on the slicing problem, they are quite sensitive on the density variations of the point cloud, as well as on the local noise in the point cloud. To relive the difficulties, three technological requirements are identified; 1) dominant point sampling, 2) avoiding local vibration, and 3) robustness on the density changes. To satisfy these requirements, we propose a new slicing algorithm which is based on a node-sphere diagram. The algorithm has been implemented and tested with various examples.

본 논문은 자동계류시스템의 선체 형상 인식을 위한 3D LiDAR 점군 정합 기법을 제안한다. 접안 환경에서는 원거리 저밀도, 근거리 시야각 제약, 단조로운 선체 기하학적 특징으로 인해 반복 최근접점(ICP) 알고리즘의 정합 성능이 저하된다. 이에 본 연구에서는 3D LiDAR 주변광 영상의 ORB 특징점과 2D영상-3D점군 인덱스 대응을 활용하여 추정된 변환행렬을 ICP의 초기값으로 이용하고, 절단 부호 거리 함수(TSDF) 모델에서 가상의 센서 시야 절두체를 정의하여 현재 시야 내 점군만을 대응쌍 탐색에 포함시킨다. 실험 결과, 제안 기법은 기존 점-대-평면 ICP 대비 정합 정확도 58% 향상을 달성하였다.

This paper proposes a 3D LiDAR point-cloud registration method for ship-hull recognition in automatic mooring systems. During berthing, iterative closest point (ICP) performance degrades due to low point density at long range, field-of-view constraints at close range, and monotonous geometric features of ship hull surfaces. To address these issues, we extract oriented FAST and rotated BRIEF (ORB) features from the 3D LiDAR ambient image and utilize 2D image-3D point-cloud index correspondences to estimate a transformation matrix, which is then used as the initial value for ICP. Additionally, we define a virtual sensor viewport frustum in the truncated signed distance function (TSDF) model to include only points within the current field of view in the correspondence search. Experimental results demonstrate that the proposed method achieves a 58% improvement in registration accuracy compared to conventional point-to-plane ICP.