http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Liu Yan-ju,Jiang Jin-gang,Miao Feng-juan,Tao Bai-rui,Zhang Hong-lie 보안공학연구지원센터 2014 International Journal of Grid and Distributed Comp Vol.7 No.5
This paper presents a fuzzy normal estimate for mass point clouds of irregular models in reconstruction. The irregular model is complex object that some part is smooth and some parts are irregular including sharp features. Therefore, we put kNN and curvature of mass point clouds to fuzzy inference system to divide the kind of point clouds and the output of FIS can determine which part of tooth point clouds belong to. For different kinds point clouds, corresponding algorithm is given. Point clouds in the smooth area are estimated normal by PCA directly and ones in other regions of thin or sharp area are estimated by checker and attach points. This method is simpler than those complex methods used on the whole point clouds directly. The experiment results show that much time is saved and surface reconstruction is very fine than PCA and WLOP.
Hogeon Seo,Sungmoon Joo 한국비파괴검사학회 2021 한국비파괴검사학회지 Vol.41 No.1
Laser scanning is a noncontact and nondestructive technique that captures the three-dimensional (3D) shape of objects as point clouds. Deep neural networks have been widely used to classify the 3D shapes of point clouds. In applying deep learning on point clouds, point cloud preprocessing is the first step. This study was conducted to analyze 3D shape classification characteristics using a deep neural network, PointNet, with a point cloud dataset, ModelNet40, for four preprocessing cases: random, scaling, zero-mean, and normalization. For each preprocessing case, the minimum and maximum coordinates of the point clouds and 3D shape classification performance are investigated. The results show that normalization preprocessing exhibits the most significant improvement in classification performance, and the zero-mean method is particularly effective. The findings indicate that proper preprocessing, such as normalization, should be performed before deep learning when the mean coordinates and scale of the point clouds differ significantly.
Hogeon Seo,Sungmoon Joo 제어로봇시스템학회 2020 제어로봇시스템학회 국제학술대회 논문집 Vol.2020 No.10
Three-dimensional (3D) laser scanning is widely used to acquire the structural information of a target as a point cloud and reconstruct its shape. Recently, deep learning has shown good performance for 3D point cloud shape classification. The preprocessing of the point cloud is a primary step of deep learning. This study presents the performance of 3D shape classification via PointNet with a point cloud dataset, ModelNet40, with respect to three preprocessing cases: Random, zero mean, and normalization. The minimum and maximum values of the point cloud are compared according to the preprocessing method. In training, the number of points as an input was 1024. In addition, the influence of two augmentation methods (i.e., resampling and zero filling) was investigated. For this, the number of points was increased to 2048. Of the 2048 points, 1024 points were used the same as in the previous experiment, while the remaining 1024 points were added by resampling or zero filling. The results show that the zero mean method is effective for deep learning and normalization is better, whereas increasing the input size with the resampling or zero filling rather degrades the performance and increases unnecessary training costs.
도심 자율주행을 위한 라이다 정지 장애물 지도 기반 위치 보정 알고리즘
노한석,이현성,이경수 사단법인 한국자동차안전학회 2022 자동차안전학회지 Vol.14 No.2
This paper presents LiDAR static obstacle map based vehicle position correction algorithm for urban autonomous driving. Real Time Kinematic (RTK) GPS is commonly used in highway automated vehicle systems. For urban automated vehicle systems, RTK GPS have some trouble in shaded area. Therefore, this paper represents a method to estimate the position of the host vehicle using AVM camera, front camera, LiDAR and low-cost GPS based on Extended Kalman Filter (EKF). Static obstacle map (STOM) is constructed only with static object based on Bayesian rule. To run the algorithm, HD map and Static obstacle reference map (STORM) must be prepared in advance. STORM is constructed by accumulating and voxelizing the static obstacle map (STOM). The algorithm consists of three main process. The first process is to acquire sensor data from low-cost GPS, AVM camera, front camera, and LiDAR. Second, low-cost GPS data is used to define initial point. Third, AVM camera, front camera, LiDAR point cloud matching to HD map and STORM is conducted using Normal Distribution Transformation (NDT) method. Third, position of the host vehicle position is corrected based on the Extended Kalman Filter (EKF).The proposed algorithm is implemented in the Linux Robot Operating System (ROS) environment and showed better performance than only lane-detection algorithm. It is expected to be more robust and accurate than raw lidar point cloud matching algorithm in autonomous driving.
Improved semantic segmentation network using normal vector guidance for LiDAR point clouds
Kim Minsung,Oh Inyoung,Yun Dongho,Ko Kwanghee 한국CDE학회 2023 Journal of computational design and engineering Vol.10 No.6
As Light Detection and Ranging (LiDAR) sensors become increasingly prevalent in the field of autonomous driving, the need for accurate semantic segmentation of three-dimensional points grows accordingly. To address this challenge, we propose a novel network model that enhances segmentation performance by utilizing normal vector information. Firstly, we present a method to improve the accuracy of normal estimation by using the intensity and reflection angles of the light emitted from the LiDAR sensor. Secondly, we introduce a novel local feature aggregation module that integrates normal vector information into the network to improve the performance of local feature extraction. The normal information is closely related to the local structure of the shape of an object, which helps the network to associate unique features with corresponding objects. We propose four different structures for local feature aggregation, evaluate them, and choose the one that shows the best performance. Experiments using the SemanticKITTI dataset demonstrate that the proposed architecture outperforms both the baseline models, RandLA-Net, and other existing methods, achieving mean intersection over union of 57.9%. Furthermore, it shows highly competitive performance compared with RandLA-Net for small and dynamic objects in a real road environment. For example, it yielded 95.2% for cars, 47.4% for bicycles, 41.0% for motorcycles, 57.4% for bicycles, and 53.2% for pedestrians.
LiDAR를 활용한 ROS 기반 ACC 인지 시스템 및 NDT-Localization 시스템에 대한 연구
김상준(Sangjun Kim),길현준(Hyeonjun Gil),최윤중(Yunjung Choi),김정하(Jungha Kim) 한국자동차공학회 2022 한국자동차공학회 학술대회 및 전시회 Vol.2022 No.11
This paper proposes ROS-Based ACC Recognition System and NDT-Localization System using LiDAR Sensor. The platform selected as the sub-controller is ERP-42, which informs the current speed, steering angle, and driving mode of the vehicle. An industrial PC is combined on the vehicle and used as a upper-Level controller. In the upper- Level controller, object detection and localization through LiDAR Sensor are published in the topic form on ROS. The vehicle performs object detection and localization through real-time communication using the corresponding topic. In conclusion, this system enables object detection and localization for ACC with one PC. By developing this study, it will be possible to build a optimized perception system using LiDAR Sensor.