Zhike Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhike Chen has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Cheng Zhou
The School of Computer Science, Guangdong Polytechnic Normal University, Guangzhou, China
2 shared publications - Guang'AN Luo
Guangdong Polytechnic Normal University
2 shared publications - Jun Liu
Guangdong Polytechnic Normal University
2 shared publications - Xinyu Wu
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences
2 shared publications - Xu Lu
Guangdong Polytechnic Normal University, Guangzhou, China
2 shared publications - Haijun Liu
The Institute of Interdisciplinary Studies, Guangdong Polytechnic Normal University, Guangzhou, China
1 shared publication
Latest Publications
- 21 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Latin America Transactions
- Vol. 24 (5) , 484-493
GASegNet: Global Self-Attention Mechanism Meets Structural Feature Fusion for Point Cloud Semantic Segmentation
With the rapid development of autonomous driving technology, semantic segmentation, as one of the key technologies contributing to the environment perception of autonomous driving systems, still suffers from a lack of connections between local features, as well as high computational consumption and an inability to meet real-time requirements. To address the above problems, this paper proposes a lightweight and efficient point cloud semantic segmentation network based on spherical projection with encoder-decoder structure. The encoder consists of a global self-attention mechanism that captures global information, as well as multi-scale convolution. This module achieves the unification of local feature extraction and global characteristic information for high-dimensional semantic information. In order to alleviate the high computational cost, a feature fusion module is introduced to enhance the compactness of the range image structure obtained from point cloud projection. The decoder utilizes bilinear interpolation to upsample multi-resolution feature maps, and introduces multiple auxiliary segmentation heads to further enhance the network's accuracy. Experiments conducted on the SemanticKITTI and SemanticPOSS datasets reveal that, in comparison to the CENet architecture, the proposed approach attains enhancements in mIoU of 4.3% and 2.6% on the respective datasets, thereby substantiating its efficacy. The code is available at GitHub: https://github.com/haifeng925/GASegNet.
- 21 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Latin America Transactions
- Vol. 24 (5) , 484-493
GASegNet: Global Self-Attention Mechanism Meets Structural Feature Fusion for Point Cloud Semantic Segmentation
With the rapid development of autonomous driving technology, semantic segmentation, as one of the key technologies contributing to the environment perception of autonomous driving systems, still suffers from a lack of connections between local features, as well as high computational consumption and an inability to meet real-time requirements. To address the above problems, this paper proposes a lightweight and efficient point cloud semantic segmentation network based on spherical projection with encoder-decoder structure. The encoder consists of a global self-attention mechanism that captures global information, as well as multi-scale convolution. This module achieves the unification of local feature extraction and global characteristic information for high-dimensional semantic information. In order to alleviate the high computational cost, a feature fusion module is introduced to enhance the compactness of the range image structure obtained from point cloud projection. The decoder utilizes bilinear interpolation to upsample multi-resolution feature maps, and introduces multiple auxiliary segmentation heads to further enhance the network's accuracy. Experiments conducted on the SemanticKITTI and SemanticPOSS datasets reveal that, in comparison to the CENet architecture, the proposed approach attains enhancements in mIoU of 4.3% and 2.6% on the respective datasets, thereby substantiating its efficacy. The code is available at GitHub: https://github.com/haifeng925/GASegNet.
- 4 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-20
RSD-SLAM: A Robust Saliency-Driven Visual SLAM System in Indoor Environments
The region of interest (ROI) with abundant and structured textures provides robust features in an indoor environment, which can effectively facilitate accurate simultaneous localization and mapping (SLAM). However, most existing visual SLAM systems generally treat ROI and non-ROI uniformly, resulting in ineffective employment of ROI. To meet this gap, we propose a robust saliency-driven visual SLAM system for indoor environments, coined RSD-SLAM. It can increase the focus on valuable ROI with the saliency maps obtained from a novel saliency prediction (SP) model. Specifically, we first design a saliency map construction method for visual SLAM enabling the SP model to accurately describe ROI, which generates the first indoor SP dataset integrating geometric, semantic, depth, and low-level visual information. Second, we develop a global stability constraint module for the SP model to enable the capability of keeping temporal consistency and illumination invariance. Third, we design a saliency map based hybrid saliency-driven mechanism to increase the focus of the system on ROI. At the front-end of the system, an adaptive feature-point extraction algorithm extracts more robust feature-points from ROI, and a saliency entropy based keyframe selection algorithm selects keyframes with the saliency value distribution of feature points. At the back-end, a dynamic weighted bundle adjustment (BA) optimization algorithm heavily weights the map points of ROI. Last, the particular focus on ROI results in a robust and accurate location. Extensive experiments, conducted on the EuRoC and TUM RGB-D datasets as well as in simulation environments, demonstrate that the proposed RSD-SLAM significantly outperforms the state-of-the-art in robustness and accuracy.
- 4 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-20
RSD-SLAM: A Robust Saliency-Driven Visual SLAM System in Indoor Environments
The region of interest (ROI) with abundant and structured textures provides robust features in an indoor environment, which can effectively facilitate accurate simultaneous localization and mapping (SLAM). However, most existing visual SLAM systems generally treat ROI and non-ROI uniformly, resulting in ineffective employment of ROI. To meet this gap, we propose a robust saliency-driven visual SLAM system for indoor environments, coined RSD-SLAM. It can increase the focus on valuable ROI with the saliency maps obtained from a novel saliency prediction (SP) model. Specifically, we first design a saliency map construction method for visual SLAM enabling the SP model to accurately describe ROI, which generates the first indoor SP dataset integrating geometric, semantic, depth, and low-level visual information. Second, we develop a global stability constraint module for the SP model to enable the capability of keeping temporal consistency and illumination invariance. Third, we design a saliency map based hybrid saliency-driven mechanism to increase the focus of the system on ROI. At the front-end of the system, an adaptive feature-point extraction algorithm extracts more robust feature-points from ROI, and a saliency entropy based keyframe selection algorithm selects keyframes with the saliency value distribution of feature points. At the back-end, a dynamic weighted bundle adjustment (BA) optimization algorithm heavily weights the map points of ROI. Last, the particular focus on ROI results in a robust and accurate location. Extensive experiments, conducted on the EuRoC and TUM RGB-D datasets as well as in simulation environments, demonstrate that the proposed RSD-SLAM significantly outperforms the state-of-the-art in robustness and accuracy.