Article
Geography, Physical
Jianwei Li, Jiawang Zhan, Ting Zhou, Virgilio A. Bento, Qianfeng Wang
Summary: This paper proposes a method for coarse registration and localization based on extracting plane features using voxels, achieving more efficient global localization and pose localization. Experimental results show that the method has high successful registration and localization rates.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Waqas Ali, Peilin Liu, Rendong Ying, Zheng Gong
Summary: This paper proposes a feature-based SLAM algorithm using 2D image projections to reduce computational cost while providing accurate and stable features. The method involves rasterizing 3D point clouds to images and applying ORB feature detector. Experimental results show significant cost reduction and accurate results.
IEEE SENSORS JOURNAL
(2021)
Article
Computer Science, Artificial Intelligence
Xiaoxia Xing, Yinghao Cai, Tao Lu, Yiping Yang, Dayong Wen
Summary: The paper introduces 3DTDesc, a data-driven descriptor that combines 2D texture and 3D geometric information for frame registration. The proposed descriptor is learned directly from color point clouds, providing efficient and robust geometric feature matching. Multi-scale 3DTDesc is also proposed to further enhance the feature matching performance, as demonstrated by extensive experimental results on challenging RGB-D datasets.
MACHINE VISION AND APPLICATIONS
(2021)
Article
Multidisciplinary Sciences
J. C. Stinville, J. M. Hestroffer, M. A. Charpagne, A. T. Polonsky, M. P. Echlin, C. J. Torbet, V. Valle, K. E. Nygren, M. P. Miller, O. Klaas, A. Loghin, I. J. Beyerlein, T. M. Pollock
Summary: The development of high-fidelity mechanical property prediction models relies on large volumes of microstructural feature data. However, spatially correlated measurements of 3D microstructure and deformation fields have been rare. This study presents a unique multi-modal dataset that combines state-of-the-art experimental techniques for 3D tomography and high-resolution deformation field measurements.
Article
Automation & Control Systems
Peizhi Shi, Qunfen Qi, Yuchu Qin, Paul J. Scott, Xiangqian Jiang
Summary: In this study, a novel deep learning approach named SsdNet is proposed to tackle the machining feature localization and recognition problem, achieving state-of-the-art performance in feature recognition and localization. The method modifies the network architecture and output of SSD, and utilizes advanced techniques to enhance recognition performance.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Computer Science, Information Systems
Shipu Diao, Yong Yang, Guanqun Cui, Yubing Chen
Summary: In this study, a novel cross-source point cloud registration technique is proposed to plan machining paths for complex cast parts by combining a point cloud slicing algorithm and a curve fitting algorithm. The results demonstrate that the proposed method significantly improves the uniformity of the surface profile and the operational effectiveness of the machining process compared to the manual method. This method may have practical applications in online inspection by UAVs and interactive robots.
COMPUTER COMMUNICATIONS
(2023)
Letter
Chemistry, Multidisciplinary
Pauline J. Kolbeck, Mihir Dass, Irina V. Martynenko, Relinde J. A. van Dijk-Moes, Kelly J. H. Brouwer, Alfons van Blaaderen, Willem Vanderlinden, Tim Liedl, Jan Lipfert
Summary: Atomic force microscopy (AFM) is a powerful technique for high-resolution imaging of molecules, macromolecular complexes, and nanoparticles. However, the shape of the AFM tip can distort the images. In this study, we use a 3D DNA origami structure as a fiducial for tip reconstruction and image correction. The fiducial has sharp steps at different heights, allowing reliable tip reconstruction with as few as ten fiducials. This fiducial enables accurate and precise AFM imaging for a wide range of applications.
Article
Computer Science, Artificial Intelligence
Yuhe Zhang, Chunhui Li, Bao Guo, Chenhao Guo, Shunli Zhang
Summary: This article introduces a novel feature descriptor named Kernel Density Descriptor (KDD), which encodes the information of the whole 3D space around the feature point through kernel density estimation and provides a strategy for selecting different matching metrics for datasets with diverse levels of resolution qualities. Experimental results demonstrate the superior performance of the KDD descriptor in terms of descriptiveness, robustness, and compactness, validating its overall superiority.
PATTERN RECOGNITION
(2021)
Article
Chemistry, Multidisciplinary
Ruiyang Sun, Enzhong Zhang, Deqiang Mu, Shijun Ji, Ziqiang Zhang, Hongwei Liu, Zheng Fu
Summary: This paper proposes a new point cloud registration method that optimizes the rough registration and precise registration stages. By improving the feature point extraction and point cloud filtering methods, and introducing the voxel concept for point cloud filtering, experimental results show noise removal rates of 95.3%, 98.6%, and 93.5%. In the precise registration stage, a method combining curvature feature and fast point feature histogram is proposed and analyzed experimentally. The analysis and verification of datasets such as Stanford bunny and free-form surface show approximately 40.16% and 36.27% reduction in error, and approximately 42.9% and 37.14% improvement in iteration times.
APPLIED SCIENCES-BASEL
(2023)
Article
Chemistry, Analytical
Xiaokai Xia, Zhiqiang Fan, Gang Xiao, Fangyue Chen, Yu Liu, Yiheng Hu
Summary: Three-dimensional point cloud registration is a widely studied problem in computer vision with various applications. Learning-based approaches, particularly attention-based models, have shown effectiveness for this task. However, using attention mechanisms in an encoder-decoder framework compromises the effectiveness of the attention module. To address this, a novel model with attention layers embedded in both the encoder and decoder stages is proposed. Experimental results on public datasets demonstrate the quality results achieved by the proposed model.
Article
Computer Science, Artificial Intelligence
Uzair Nadeem, Mohammed Bennamoun, Roberto Togneri, Ferdous Sohel, Aref Miri Rekavandi, Farid Boussaid
Summary: This paper presents a novel approach for cross-domain descriptor matching between 2D and 3D modalities. The proposed technique allows localization of 2D images in any type of 3D point cloud without constraints on its nature or acquisition mechanism.
PATTERN RECOGNITION
(2023)
Article
Optics
Hao Ma, De-Yu Yin, Jing-Bin Liu, Rui-Zhi Chen
Summary: In this paper, a fully unsupervised feature matching method based on multi-resolution voxel model is proposed. Features are extracted from the voxel model using a convolutional auto-encoder (CAE), and the initial corresponding features are determined based on Euclidean distance. Refined matches are obtained through RANSAC algorithm. Experiments show that this method outperforms state-of-the-art techniques in average matching inlier ratio by almost 20-30%.
OPTICS AND LASER TECHNOLOGY
(2022)
Article
Environmental Sciences
Yeping Peng, Shengdong Lin, Hongkun Wu, Guangzhong Cao
Summary: Three-dimensional (3D) reconstruction is crucial for visualizing and monitoring plant growth, but inspecting tall plants using single-camera systems is challenging. This study proposes a combination of low-altitude remote sensing and a terrestrial capture platform to obtain structural data of trees. A registration method based on FPFH is introduced to align tree point clouds from different sensors, with initial correspondences calculated using Bhattacharyya distance and reliable matches identified through random sample consensus. Real-world experiments validate the effectiveness of the proposed method. The root-mean-square error of the proposed method is 0.35% and 1.18% of SAC-IA and SAC-IA + ICP, respectively. The proposed method is valuable for plant phenotype extraction, monitoring, and analysis.
Article
Computer Science, Interdisciplinary Applications
Wei Fan, Lianyu Zheng, Wei Ji, Xun Xu, Yuqian Lu, Lihui Wang
Summary: In order to ensure the final assembly accuracy of large aircraft components, an adaptive positioning method is proposed in this paper, which integrates comprehensive engineering constraints to improve machining efficiency and accuracy by finishing machining the assembly interface of large components and using an optimization algorithm.
ROBOTICS AND COMPUTER-INTEGRATED MANUFACTURING
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Shixin Yang, Haojiang Li, Shuchao Chen, Wenjie Huang, Demin Liu, Guangying Ruan, Qiangyang Huang, Qiong Gong, Lizhi Liu, Hongbo Chen
Summary: We propose a multiscale feature fusion registration based on deep learning to achieve the accurate registration and fusion of head magnetic resonance imaging (MRI) and solve the problem that general registration methods cannot handle the complex spatial information and position information of head MRI.