Zhentao Guo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhentao Guo has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Guiyu Zhao
National Key Laboratory of Autonomous Intelligent Unmanned Systems, School of Automation, Beijing Institute of Technology, Beijing, China
2 shared publications - Hongbin Ma
School of Automation, Beijing Institute of Technology, Beijing 100081, China
2 shared publications - Tongkun Guan
Shanghai Jiao Tong University
2 shared publications - Zining Wang
Meituan
2 shared publications - Boming Chen1 shared publication
- Chao Jin
Chinese Academy of Sciences
1 shared publication
Latest Publications
- 3 October 2026
- Published by Springer Nature
- p. 712-730
ViTexQA: A Multi-frame Temporal Perception Dataset for Video Text Question Answering
- 3 October 2026
- Published by Springer Nature
- p. 712-730
ViTexQA: A Multi-frame Temporal Perception Dataset for Video Text Question Answering
- 18 September 2026
- Published by Springer Nature
- p. 515-534
AdaThinking- : One-Token Entropy Regulation for Adaptive Thinking
- 18 September 2026
- Published by Springer Nature
- p. 515-534
AdaThinking- : One-Token Entropy Regulation for Adaptive Thinking
- 13 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-15
GOR-PCR: Graph-Based Outlier Removal via Global-to-Local Consistency for Point Cloud Registration
In point cloud registration, the point cloud with much noise or low overlap will cause more outliers, consequently leading to registration failure. This paper introduces a graph-based outlier removal method with global-to-local consistency for robust point cloud registration. To address the issue of global ambiguity inherent in second-order consistency, we introduce the concept of global-to-local outlier removal. This approach preserves the outlier filtering capability of global consistency while mitigating the ambiguity associated with global consistency. Furthermore, local consistency enhances the detection of local overlapping regions. To bolster the robustness of local consistency in identifying outliers, we propose a dual criterion based on both distance and angle. Finally, robust global registration is achieved by overlap-aware hypothesis verification. Our training-free method outperforms other outlier removal methods, achieving state-of-the-art results of 96.80% RR and 93.79% IR on the 3DMatch dataset. In the outdoor KITTI dataset, our method demonstrates the capability to achieve robust and accurate registration using the less informative FPFH descriptor, attaining the highest RR of 99.82% and the lowest RE of 8.11. Additionally, this method can be effectively integrated with learning-based point cloud registration, which significantly improves their registration recall without introducing much time and space overhead. Moreover, our method improves the feature-based methods by 1.0 ∼ 29.9 pp on RR and 8.0 ∼ 51.3 pp on IR. The codes will be released soon.
- 13 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-15
GOR-PCR: Graph-Based Outlier Removal via Global-to-Local Consistency for Point Cloud Registration
In point cloud registration, the point cloud with much noise or low overlap will cause more outliers, consequently leading to registration failure. This paper introduces a graph-based outlier removal method with global-to-local consistency for robust point cloud registration. To address the issue of global ambiguity inherent in second-order consistency, we introduce the concept of global-to-local outlier removal. This approach preserves the outlier filtering capability of global consistency while mitigating the ambiguity associated with global consistency. Furthermore, local consistency enhances the detection of local overlapping regions. To bolster the robustness of local consistency in identifying outliers, we propose a dual criterion based on both distance and angle. Finally, robust global registration is achieved by overlap-aware hypothesis verification. Our training-free method outperforms other outlier removal methods, achieving state-of-the-art results of 96.80% RR and 93.79% IR on the 3DMatch dataset. In the outdoor KITTI dataset, our method demonstrates the capability to achieve robust and accurate registration using the less informative FPFH descriptor, attaining the highest RR of 99.82% and the lowest RE of 8.11. Additionally, this method can be effectively integrated with learning-based point cloud registration, which significantly improves their registration recall without introducing much time and space overhead. Moreover, our method improves the feature-based methods by 1.0 ∼ 29.9 pp on RR and 8.0 ∼ 51.3 pp on IR. The codes will be released soon.
- 13 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-16
SphereNet: Learning a Noise-Robust and General Descriptor for Point Cloud Registration
Point cloud registration aims to estimate a transformation that aligns point clouds collected from different perspectives. In learning-based point cloud registration, a robust descriptor is crucial for achieving high-accuracy registration. However, most existing methods are susceptible to noise and demonstrate poor generalization ability when applied to unseen datasets. Motivated by this, we introduce SphereNet to learn a noise-robust and unseen-general descriptor for point cloud registration. In our method, first, the spheroid generator builds a geometric domain based on spherical voxelization (SV) to encode geometric information. Then, the spherical interpolation of the sphere is introduced to realize robustness against noise. Finally, a new spherical convolutional neural network (CNN) with spherical integrity padding completes the extraction of descriptors, which reduces the loss of features and fully captures the geometric features. To evaluate our methods, a new benchmark 3DMatch-noise with strong noise is introduced. Extensive experiments are carried out on both indoor and outdoor datasets. Our results demonstrate that SphereNet achieves an increase in feature-matching recall of more than 25 percentage points (pp) on 3DMatch-noise under high-intensity noise. Moreover, SphereNet establishes a new state-of-the-art performance on the 3DMatch and 3DLoMatch benchmarks, achieving 93.5% and 75.6% registration recall (RR), respectively. Furthermore, SphereNet exhibits superior generalization ability on unseen datasets.
- 13 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-16
SphereNet: Learning a Noise-Robust and General Descriptor for Point Cloud Registration
Point cloud registration aims to estimate a transformation that aligns point clouds collected from different perspectives. In learning-based point cloud registration, a robust descriptor is crucial for achieving high-accuracy registration. However, most existing methods are susceptible to noise and demonstrate poor generalization ability when applied to unseen datasets. Motivated by this, we introduce SphereNet to learn a noise-robust and unseen-general descriptor for point cloud registration. In our method, first, the spheroid generator builds a geometric domain based on spherical voxelization (SV) to encode geometric information. Then, the spherical interpolation of the sphere is introduced to realize robustness against noise. Finally, a new spherical convolutional neural network (CNN) with spherical integrity padding completes the extraction of descriptors, which reduces the loss of features and fully captures the geometric features. To evaluate our methods, a new benchmark 3DMatch-noise with strong noise is introduced. Extensive experiments are carried out on both indoor and outdoor datasets. Our results demonstrate that SphereNet achieves an increase in feature-matching recall of more than 25 percentage points (pp) on 3DMatch-noise under high-intensity noise. Moreover, SphereNet establishes a new state-of-the-art performance on the 3DMatch and 3DLoMatch benchmarks, achieving 93.5% and 75.6% registration recall (RR), respectively. Furthermore, SphereNet exhibits superior generalization ability on unseen datasets.