A5084787849- Robotics and Sensor-Based Localization
- 3D Surveying and Cultural Heritage
- 3D Shape Modeling and Analysis
- Advanced Image and Video Retrieval Techniques
- Image Processing and 3D Reconstruction
- Advanced Neural Network Applications
- Remote Sensing and LiDAR Applications
- Image and Object Detection Techniques
- Advanced Vision and Imaging
- Medical Image Segmentation Techniques
- Advanced Image Fusion Techniques
- Infrared Target Detection Methodologies
- Optical measurement and interference techniques
- Advanced Optical Sensing Technologies
- Automated Road and Building Extraction
- Advanced Image Processing Techniques
- Image Retrieval and Classification Techniques
- Robot Manipulation and Learning
- Autonomous Vehicle Technology and Safety
- Vehicle License Plate Recognition
- Underwater Vehicles and Communication Systems
- Video Surveillance and Tracking Methods
- Industrial Vision Systems and Defect Detection
- Hand Gesture Recognition Systems
- Digital Holography and Microscopy
Chang'an University
2019-2025
Shenzhen University
2025
Northwestern Polytechnical University
2019-2020
Huazhong University of Science and Technology
2016-2019
This article presents an efficient and robust estimator called compatibility-guided sampling consensus (CG-SAC) to achieve accurate 3-D point cloud registration. For correspondence-based registration methods, the random sample (RANSAC) is served as a de facto solution for rigid transformation estimation from number of feature correspondences. Unfortunately, RANSAC still suffers two major limitations. First, it generates hypothesis with at least three samples desires very large iterations...
Six-degree-of-freedom (6-DOF) pose estimation from feature correspondences remains a popular and robust approach for 3-D registration. However, heavy outliers that existed in the initial correspondence set great challenge to this problem. This article presents simple yet effective estimator called SAmple Consensus by sampling COmpatibility Triangles graphs (SAC-COT) 6-DOF The key novelty is guided three-point approach. It based on novel sample representation, i.e., Triangle (COT). We first...
This paper focuses on developing efficient and robust evaluation metrics for RANSAC hypotheses to achieve accurate 3D rigid registration. Estimating six-degree-of-freedom (6-DoF) pose from feature correspondences remains a popular approach registration, where random sample consensus (RANSAC) is well-known solution this problem. However, existing are either time-consuming or sensitive common nuisances, parameter variations, different application scenarios, resulting in performance...
This article presents a simple yet effective method for 3-D correspondence selection and point cloud registration. It first models the initial set as graph with nodes representing correspondences edges connecting geometrically compatible nodes. Such graphs offer either loose or tight geometric constraints judging correctness of correspondence, e.g., edges, loops, cliques. Then, we render these dynamic voters to judge node. More specifically, develop loose–tight voting (LT-GV) that employs...
3D scene completion (SC) has made progress in the last three years. From application of mobile robot system, SC should support downstream task (i.e. mapping or perception), instead only predicting completed scenes. However, as low-cost few-beam LiDAR is widely applied robot, gap between and tasks large. To generate high quality result, bottleneck lies triple sparsity input, ground truth (GT) occupancy, GT foreground. deal with sparsity, we present an extreme sparse network (ESC-Net). At...
Abstract Correspondence-based six-degree-of-freedom (6-DoF) pose estimation remains a mainstream solution for 3D point cloud registration. However, the heavy outliers great challenges to this problem. In paper, we propose random sample consensus (RANSAC) variant based on sampling locally and hypothesis globally (SLHG) 6-DoF The key novelties are efficient by guiding process accurate generating hypotheses with global information. SLHG first generates correspondence subset via compatibility...
Estimating an accurate six-degree-of-freedom (6-DoF) pose from correspondences with outliers remains a critical issue to 3D rigid registration. Random sample consensus (RANSAC) and its variants are popular solutions this problem. Although there have been number of RANSAC-fashion estimators, two issues remain unsolved. First, it is unclear which estimator more appropriate particular application. Second, the impacts different sampling strategies, hypothesis generation methods, evaluation...
Local geometric descriptors remain an essential component for 3D rigid data matching and fusion. The devise of a rotational invariant local descriptor usually consists two steps: reference frame (LRF) construction feature representation. Existing evaluation efforts have mainly been paid on the LRF or overall descriptor, yet quantitative comparison representations remains unexplored. This paper fills this gap by comprehensively evaluating nine state-of-the-art representations. Our is ground...
Three dimensional (3D) object detection with an optical camera and light ranging (LiDAR) is essential task in the field of mobile robot autonomous driving. The current 3D method based on deep learning data-hungry. Recently, semi-supervised (SSOD-3D) has emerged as a technique to alleviate shortage labeled samples. However, it still challenging problem for SSOD-3D learn from noisy pseudo labels. In this paper, dynamically filter unreliable labels, we first introduce self-paced SPSL-3D. It...
This paper presents a fast deblurring algorithm to remove camera motion blur from single photograph using built-in gyroscopes and strong edge prediction. An inaccurate kernel or point spread function (PSF) usually leads an unsatisfying restored result. Hence, we propose robust three-phase method for accurate PSF estimation. In the first stage, utilize embedded compute coarse version of camera's angular velocity during exposure. order reduce execution time later modification, introduce patch...
Feature matching for 3D point clouds is a fundamental yet challenging problem in remote sensing and computer vision. However, due to number of nuisances, the initial feature correspondences generated by local keypoint descriptors may contain many outliers (incorrect correspondences). To remove outliers, this paper presents robust method called progressive consistency voting (PCV). PCV aims at assigning reliable confidence score each correspondence such that reasonable can be achieved simply...