Baozhi Jia
Research Subjects
Baozhi Jia has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Jiankun Wang
Southern University of Science and Technology
3 shared publications - Hanchao Wang
Reconova Co., Ltd., Xiamen, Fujian, China
2 shared publications - Bi-Ke Zhu
Southern University of Science and Technology
1 shared publication - Bingyi Xia
SUSTech, Shenzhen, China
1 shared publication - Chaoyou Fu
Nanjing University, Nanjing, China
1 shared publication - Dingyuan Tang
Southwest Jiaotong University
1 shared publication
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 13686-13694
Tango: Taming Visual Signals for Efficient Video Large Language Models
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 13686-13694
Open AccessTango: Taming Visual Signals for Efficient Video Large Language Models
- 25 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (5) , 6186-6193
Learning Robot Visual Navigation in Crowds via Intention-Aware Scene Representations
Robot crowd navigation requires the ability to infer human intentions while accounting for the structural constraints of the environment. Currently, deep reinforcement learning (DRL) provides a promising method for learning navigation policies that understand human intentions. However, most of them rely on limited scene representations, treating pedestrians as simple 2D points and ignoring rich visual cues from both humans and the environment. To address this issue, iCrowdNav, a novel visual crowd navigation method with intention-aware scene representations, is introduced to encode behavioral and structural context from egocentric visual observations. Our method employs two key components: a spatio-temporal encoder for extracting occupancy features of the scene, and Intent-Interact Former (I$^{2}$Former), an attention-based module that encodes human poses to infer pedestrians' motion intentions. These features are integrated into a compact state embedding that supports effective DRL policy training. Extensive experiments show that our method achieves superior performance over baselines, and real-world deployment demonstrates vision-based crowd navigation.
- 25 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (5) , 6186-6193
Learning Robot Visual Navigation in Crowds via Intention-Aware Scene Representations
Robot crowd navigation requires the ability to infer human intentions while accounting for the structural constraints of the environment. Currently, deep reinforcement learning (DRL) provides a promising method for learning navigation policies that understand human intentions. However, most of them rely on limited scene representations, treating pedestrians as simple 2D points and ignoring rich visual cues from both humans and the environment. To address this issue, iCrowdNav, a novel visual crowd navigation method with intention-aware scene representations, is introduced to encode behavioral and structural context from egocentric visual observations. Our method employs two key components: a spatio-temporal encoder for extracting occupancy features of the scene, and Intent-Interact Former (I$^{2}$Former), an attention-based module that encodes human poses to infer pedestrians' motion intentions. These features are integrated into a compact state embedding that supports effective DRL policy training. Extensive experiments show that our method achieves superior performance over baselines, and real-world deployment demonstrates vision-based crowd navigation.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 19503-19515
A Vision-Driven Closed-Loop Framework for High-Precision Robotic Baggage Handling
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 19503-19515
A Vision-Driven Closed-Loop Framework for High-Precision Robotic Baggage Handling
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Industry Applications
- Vol. 7, 1193-1206
Communication-Free Efficiency Optimization of a Wireless Power Transfer System With Planar Misalignment Tolerance
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Industry Applications
- Vol. 7, 1193-1206
Open AccessCommunication-Free Efficiency Optimization of a Wireless Power Transfer System With Planar Misalignment Tolerance
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 14837-14851
LSTP-Nav: Lightweight Spatiotemporal Policy for Map-Free Multi-Agent Navigation With LiDAR
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 14837-14851
LSTP-Nav: Lightweight Spatiotemporal Policy for Map-Free Multi-Agent Navigation With LiDAR
- 16 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-15
A Single-Frequency Autofocusing Method Based on Conditional Diffusion Model for Millimeter-Wave Near-Range Imaging
Active wideband millimeter-wave (MMW) systems enable safe and accurate 3-D microwave imaging for various applications, particularly in security screening, while intricate design and implementation are often employed to acquire wideband measurements. Potential imaging methods utilizing single-frequency measurements can be designed to alleviate the hardware requirements. However, applying conventional near-range imaging algorithms directly to single-frequency measurements leads to defocusing issues and degraded imaging results due to the lack of range resolution. Although previous studies have employed autofocusing methods to achieve focused results from defocused images, further enhancement is required for the imaging quality of extended targets. An innovative method based on the conditional diffusion model, named single-frequency autofocusing diffusion model (SFADiff), is proposed in this paper to address the defocusing problem in single-frequency imaging and reconstruct high-quality images effectively. The single-frequency autofocusing encoder network (SFAENet) is first developed to convert 3-D defocused images into 2-D encoded images. Subsequently, SFADiff utilizes these encoded images as conditions guiding image generation towards producing high-quality focused images. Comprehensive experiments are conducted to demonstrate that SFADiff exhibits exceptional performance in terms of image quality, offering a viable solution for achieving single-frequency imaging through autofocusing.
- 16 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-15
A Single-Frequency Autofocusing Method Based on Conditional Diffusion Model for Millimeter-Wave Near-Range Imaging
Active wideband millimeter-wave (MMW) systems enable safe and accurate 3-D microwave imaging for various applications, particularly in security screening, while intricate design and implementation are often employed to acquire wideband measurements. Potential imaging methods utilizing single-frequency measurements can be designed to alleviate the hardware requirements. However, applying conventional near-range imaging algorithms directly to single-frequency measurements leads to defocusing issues and degraded imaging results due to the lack of range resolution. Although previous studies have employed autofocusing methods to achieve focused results from defocused images, further enhancement is required for the imaging quality of extended targets. An innovative method based on the conditional diffusion model, named single-frequency autofocusing diffusion model (SFADiff), is proposed in this paper to address the defocusing problem in single-frequency imaging and reconstruct high-quality images effectively. The single-frequency autofocusing encoder network (SFAENet) is first developed to convert 3-D defocused images into 2-D encoded images. Subsequently, SFADiff utilizes these encoded images as conditions guiding image generation towards producing high-quality focused images. Comprehensive experiments are conducted to demonstrate that SFADiff exhibits exceptional performance in terms of image quality, offering a viable solution for achieving single-frequency imaging through autofocusing.