Zitai Zhou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zitai Zhou has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Jie Zhang
University of Science and Technology Beijing
2 shared publications - Junhua Sun
School of Instrumentation and Optoelectronic Engineering, Beihang University, Beijing, China
2 shared publications
Latest Publications
- 1 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (2) , 2812-2821
Center-Aware Instance Segmentation for Close Objects in Industrial 3-D Point Cloud Scenes
Segmenting small-scale and close 3-D objects from a complicated 3-D point cloud scene is a crucial yet challenging task for industrial scene understanding. Previous 3-D instance segmentation work (such as PointGroup and SoftGroup) suffers from limited accuracy when dealing with objects in close proximity. This article proposes a center-aware 3-D instance segmentation framework that performs two cooperative stages, including bottom-up center-guided instance proposal generation and top-down center-aware instance mask refinement, to jointly address the challenges of segmenting close objects. In the first stage, the object clusters are first segmented from the cluttered background. We propose to determine the center of each object by predicting its k -nearest neighbors with their respective offsets, which enables an instance proposal generated for each object. In the second stage, center-based relative position encoding is performed on all points in each instance proposal. We concatenate the explicitly encoded geometric features (low-level) with high-level semantic features and construct the center-aware feature representation for instance mask prediction. We conducted experiments on a real-world industrial point cloud dataset that involves the important aero-engine component—close clamps in the cluttered scenario. The results show that our proposed method achieves state-of-the-art overall 3-D instance segmentation precision with $\mathbf{AP}_{\mathbf{avg}}$ of 0.905 and $\mathbf{AP}_{\mathbf{90}}$ of 0.736.
- 1 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (2) , 2812-2821
Center-Aware Instance Segmentation for Close Objects in Industrial 3-D Point Cloud Scenes
Segmenting small-scale and close 3-D objects from a complicated 3-D point cloud scene is a crucial yet challenging task for industrial scene understanding. Previous 3-D instance segmentation work (such as PointGroup and SoftGroup) suffers from limited accuracy when dealing with objects in close proximity. This article proposes a center-aware 3-D instance segmentation framework that performs two cooperative stages, including bottom-up center-guided instance proposal generation and top-down center-aware instance mask refinement, to jointly address the challenges of segmenting close objects. In the first stage, the object clusters are first segmented from the cluttered background. We propose to determine the center of each object by predicting its k -nearest neighbors with their respective offsets, which enables an instance proposal generated for each object. In the second stage, center-based relative position encoding is performed on all points in each instance proposal. We concatenate the explicitly encoded geometric features (low-level) with high-level semantic features and construct the center-aware feature representation for instance mask prediction. We conducted experiments on a real-world industrial point cloud dataset that involves the important aero-engine component—close clamps in the cluttered scenario. The results show that our proposed method achieves state-of-the-art overall 3-D instance segmentation precision with $\mathbf{AP}_{\mathbf{avg}}$ of 0.905 and $\mathbf{AP}_{\mathbf{90}}$ of 0.736.
- 21 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 72 (00189456) , 1-12
Three-Dimensional Segmentation and Global Clearance Analysis for Free-Bent Pipelines in Point-Cloud Scenarios
Detecting free-bent pipelines and analyzing the global-range clearance in cluttered industrial scenarios is a significant yet challenging task in pipeline inspection. Most of the previous work focused on local ranges or specific pipeline components instead of covering a wide length range. This paper proposes a comprehensive pipeline segmentation and clearance analysis framework that combines a deep pipeline feature learning module for semantic pipeline segmentation and a pipeline primitive growing strategy for measuring the clearance along the length range in a point cloud scenario. We propose to learn enhanced high-level pipeline features by incorporating an important geometric cue concerning cylindrical axial consistency, which benefits better pipeline segmentation from the point cloud background. Then, each pipeline instance is modeled as a sequence of 3D cylinder primitives. We establish a geometric model to measure the minimum panoramic clearance of each cylinder segment in the scene. Then, the pipeline clearance is tracked along the centerline in a segment sequential growing mode. Experiments were carried out on real-world industrial point cloud scenarios. The proposed pipeline semantic segmentation achieves state-of-the-art performance. The clearance analysis module reaches a mean accuracy of 0.018 mm. Our method is applicable to various free-bent pipelines and robust to cluttered backgrounds.
- 21 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 72 (00189456) , 1-12
Three-Dimensional Segmentation and Global Clearance Analysis for Free-Bent Pipelines in Point-Cloud Scenarios
Detecting free-bent pipelines and analyzing the global-range clearance in cluttered industrial scenarios is a significant yet challenging task in pipeline inspection. Most of the previous work focused on local ranges or specific pipeline components instead of covering a wide length range. This paper proposes a comprehensive pipeline segmentation and clearance analysis framework that combines a deep pipeline feature learning module for semantic pipeline segmentation and a pipeline primitive growing strategy for measuring the clearance along the length range in a point cloud scenario. We propose to learn enhanced high-level pipeline features by incorporating an important geometric cue concerning cylindrical axial consistency, which benefits better pipeline segmentation from the point cloud background. Then, each pipeline instance is modeled as a sequence of 3D cylinder primitives. We establish a geometric model to measure the minimum panoramic clearance of each cylinder segment in the scene. Then, the pipeline clearance is tracked along the centerline in a segment sequential growing mode. Experiments were carried out on real-world industrial point cloud scenarios. The proposed pipeline semantic segmentation achieves state-of-the-art performance. The clearance analysis module reaches a mean accuracy of 0.018 mm. Our method is applicable to various free-bent pipelines and robust to cluttered backgrounds.