Chengzhong Wu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Chengzhong Wu has a total of 53 co-authors. Here are the top co-authors by number of shared publications:
- Yaonan Wang
Hunan University
6 shared publications - Mingtao Feng
Hunan University
4 shared publications - Jianan Huang
Hunan University
3 shared publications - Xuebing Liu
College of Electrical and Information Engineering, Hunan University, Changsha, China
3 shared publications - Ang Li
School of Information Engineering, Chang’an University, Xi’an 710064, China
2 shared publications - Congxuan Zhang
Nanchang Hangkong University
2 shared publications
Latest Publications
- Open Access
- 5 December 2025
- Published by Springer Nature in Journal of Intelligent & Robotic Systems
- Vol. 112 (1)
Dynamic Speed and Separation Monitoring for Human-Robot Collaboration Based on Binocular Vision
- 5 December 2025
- Published by Springer Nature in Journal of Intelligent & Robotic Systems
- Vol. 112 (1)
Open AccessDynamic Speed and Separation Monitoring for Human-Robot Collaboration Based on Binocular Vision
- 2 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23 (15455955) , 417-432
SGTP-Net: Semantic Guidance and Texture Priors-Based Dual-Branch Segmentation Network for Surface Defect Detection
Deep learning-based surface defect segmentation approaches have shown promising performance in recent years. However, segmenting defects with complex shapes, large variations in size, and weakly textured defects with indistinct characteristics still poses significant challenges. In this article, a novel semantic guidance and texture priors based dual-branch surface defect segmentation network (SGTP-Net) is proposed for those issues. Firstly, we construct a feature extraction network combines semantic and texture branches. The semantic branch establishes global contextual relationships, while the texture branch captures local features of defects, this dual-branch ensured the network to extract features from various complex defects. Secondly, we design a feature fusion strategy based on semantic guidance and texture priors. The semantic information is used to guides the output of texture branch. After that, the guided texture information provides valuable edge texture priors for each layers output in semantic branch. The two branches mutually guide each other for improving ability of weak textures feature extraction. Finally, we run our method on the NEU-Seg, MT-Defect and MSD datasets to conduct a comprehensive comparison with some state-of-the-art general object segmentation models and specialized surface defect segmentation methods. The experimental results show that our SGTP-Net performs well in surface defect detection, offering excellent semantic segmentation accuracy and exhibiting good stability and robustness in detecting various surface defects.
- 2 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23 (15455955) , 417-432
SGTP-Net: Semantic Guidance and Texture Priors-Based Dual-Branch Segmentation Network for Surface Defect Detection
Deep learning-based surface defect segmentation approaches have shown promising performance in recent years. However, segmenting defects with complex shapes, large variations in size, and weakly textured defects with indistinct characteristics still poses significant challenges. In this article, a novel semantic guidance and texture priors based dual-branch surface defect segmentation network (SGTP-Net) is proposed for those issues. Firstly, we construct a feature extraction network combines semantic and texture branches. The semantic branch establishes global contextual relationships, while the texture branch captures local features of defects, this dual-branch ensured the network to extract features from various complex defects. Secondly, we design a feature fusion strategy based on semantic guidance and texture priors. The semantic information is used to guides the output of texture branch. After that, the guided texture information provides valuable edge texture priors for each layers output in semantic branch. The two branches mutually guide each other for improving ability of weak textures feature extraction. Finally, we run our method on the NEU-Seg, MT-Defect and MSD datasets to conduct a comprehensive comparison with some state-of-the-art general object segmentation models and specialized surface defect segmentation methods. The experimental results show that our SGTP-Net performs well in surface defect detection, offering excellent semantic segmentation accuracy and exhibiting good stability and robustness in detecting various surface defects.
- 17 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (12) , 12620-12627
CFPGrasp: Coarse-to-Fine Prediction Network for 6-DoF Robotic Grasping With Improved Grasp Representation
Grasping objects plays a crucial role in robotic manipulation within industrial applications, with the primary goal being the prediction of the robot grasp pose for each target object. However, accurately predicting the grasp pose is challenging due to diverse object shapes, appearances, and sizes. Moreover, the inherent complexity of grasp parameters, together with the existence of multiple valid grasps per object, imposes substantial demands on reasoning. To address these challenges, we first introduce a physics-informed contact–rotation grasp representation that simplifies the pose space, and then adopt a coarse-to-fine prediction strategy that proposes contact candidates globally and refines their rotation and quality locally. Accordingly, we propose a coarse-to-fine prediction network for 6-DoF grasping prediction in this paper, which first detects the valid contact points of grasp pose in the global scene, then predicts its corresponding rotation angle and grasp quality in the local region. The evaluation results on simulation and real-world grasp datasets demonstrate the advantages and effectiveness of the proposed approach. We also conduct a practical grasping experiment to show its generalization performance in real-world robotic grasping.
- 17 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (12) , 12620-12627
CFPGrasp: Coarse-to-Fine Prediction Network for 6-DoF Robotic Grasping With Improved Grasp Representation
Grasping objects plays a crucial role in robotic manipulation within industrial applications, with the primary goal being the prediction of the robot grasp pose for each target object. However, accurately predicting the grasp pose is challenging due to diverse object shapes, appearances, and sizes. Moreover, the inherent complexity of grasp parameters, together with the existence of multiple valid grasps per object, imposes substantial demands on reasoning. To address these challenges, we first introduce a physics-informed contact–rotation grasp representation that simplifies the pose space, and then adopt a coarse-to-fine prediction strategy that proposes contact candidates globally and refines their rotation and quality locally. Accordingly, we propose a coarse-to-fine prediction network for 6-DoF grasping prediction in this paper, which first detects the valid contact points of grasp pose in the global scene, then predicts its corresponding rotation angle and grasp quality in the local region. The evaluation results on simulation and real-world grasp datasets demonstrate the advantages and effectiveness of the proposed approach. We also conduct a practical grasping experiment to show its generalization performance in real-world robotic grasping.
- 28 February 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of High Speed Electronics and Systems
- Vol. 35 (03)
Image-Enhanced Asphalt Crack Detection Research
The main environmental factors that interfere with asphalt pavement crack detection include shadows from ambient light of different intensities, trees, signboards, railings, etc. Traditional crack detection methods usually eliminate the effect of light during pre-processing or improve the recognition results by local consistency and light normalization during image segmentation. However, the current methods can only improve the image brightness uniformity and cannot completely eliminate the effect of light, so the usual methods are only effective when the image brightness uniformity is good. So, this study puts out an approach to improving images. Machine learning is used to suggest a method for detecting cracks in asphalt pavement, with the help of an attention mechanism, using photos of real roads as the experimental dataset. Experiments show that the method can preprocess image data well and enhance the robustness of training in machine learning structures. Test results show that our method can be well applied to practical testing work.
- 28 February 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of High Speed Electronics and Systems
- Vol. 35 (03)
Image-Enhanced Asphalt Crack Detection Research
The main environmental factors that interfere with asphalt pavement crack detection include shadows from ambient light of different intensities, trees, signboards, railings, etc. Traditional crack detection methods usually eliminate the effect of light during pre-processing or improve the recognition results by local consistency and light normalization during image segmentation. However, the current methods can only improve the image brightness uniformity and cannot completely eliminate the effect of light, so the usual methods are only effective when the image brightness uniformity is good. So, this study puts out an approach to improving images. Machine learning is used to suggest a method for detecting cracks in asphalt pavement, with the help of an attention mechanism, using photos of real roads as the experimental dataset. Experiments show that the method can preprocess image data well and enhance the robustness of training in machine learning structures. Test results show that our method can be well applied to practical testing work.
- 4 February 2025
- Published by Elsevier in Optics and Lasers in Engineering
- Vol. 187
Fluorescence molecular tomography via score-based generative model
- 4 February 2025
- Published by Elsevier in Optics and Lasers in Engineering
- Vol. 187
Fluorescence molecular tomography via score-based generative model
- 24 January 2025
- Published by Springer Nature
- p. 381-396
Free-Form Instruction Guided Robotic Navigation Path Planning with Large Vision-Language Model
- 24 January 2025
- Published by Springer Nature
- p. 381-396
Free-Form Instruction Guided Robotic Navigation Path Planning with Large Vision-Language Model
- 1 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (12) , 11345-11352
A Fast and Accurate Visual Inertial Odometry Using Hybrid Point-Line Features
Mainstream visual-inertial SLAM systems use point features for motion estimation and localization. However, point features do not perform well in scenes such as weak texture and motion blur. Therefore, the introduction of line features has received a lot of attention. In this letter, we propose a point-line based real-time monocular visual inertial odometry. Aiming at the problem that most of the current works do not fully utilize the line feature properties, we derive the point-line based hybrid Multi-State Constraint Kalman Filter (hybrid MSCKF) in detail. To further improve the line feature initialization accuracy, we propose a two-step line triangulation method. Since filter-based methods are susceptible to visual outliers, we also propose a redundant line feature removal strategy suitable for the filtering framework. According to the experimental results in EuRoC data set and real environment, the proposed algorithm outperforms other state-of-the-art algorithms in accuracy and real-time performance.
- 1 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (12) , 11345-11352
A Fast and Accurate Visual Inertial Odometry Using Hybrid Point-Line Features
Mainstream visual-inertial SLAM systems use point features for motion estimation and localization. However, point features do not perform well in scenes such as weak texture and motion blur. Therefore, the introduction of line features has received a lot of attention. In this letter, we propose a point-line based real-time monocular visual inertial odometry. Aiming at the problem that most of the current works do not fully utilize the line feature properties, we derive the point-line based hybrid Multi-State Constraint Kalman Filter (hybrid MSCKF) in detail. To further improve the line feature initialization accuracy, we propose a two-step line triangulation method. Since filter-based methods are susceptible to visual outliers, we also propose a redundant line feature removal strategy suitable for the filtering framework. According to the experimental results in EuRoC data set and real environment, the proposed algorithm outperforms other state-of-the-art algorithms in accuracy and real-time performance.
- 23 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (9) , 11127-11138
PoseDiffusion: A Coarse-to-Fine Framework for Unseen Object 6-DoF Pose Estimation
Accurately estimating the six-degrees of freedom (DoF) pose of unseen objects is crucial for successful robotic manipulation in industrial automation. Some existing methods for this task rely on prior knowledge of individual objects, i.e., the model must be trained on the exact object instance or object category. Others perform unseen object pose estimation but are limited in their feature learning and pose refinement ability. To address these problems, we propose an unseen object pose estimation method that follows a coarse-to-fine framework and leverages the powerful learning ability of diffusion models. We introduce a diffusion model for generating object poses, and conduct a comparison between the generated poses and the original pose to determine the optimal one. We design a novel pose estimation module to provide coarse poses for the PoseDiffusion. This module comprises two feature extraction modules that extract global and masked features. In addition, we propose a strategy to estimate the pose by comparing the similarity between rendered and query poses. The renderings of an unseen object from various viewpoints are generated from its computer-aided design (CAD) model. Our method requires a CAD model of the unseen object only during inference, a scenario well suited to industrial applications. Experimental evaluation on benchmark datasets demonstrates that the proposed framework outperforms existing approaches, achieving state-of-the-art performance in six-DoF object pose estimation.
- 23 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 20 (9) , 11127-11138
PoseDiffusion: A Coarse-to-Fine Framework for Unseen Object 6-DoF Pose Estimation
Accurately estimating the six-degrees of freedom (DoF) pose of unseen objects is crucial for successful robotic manipulation in industrial automation. Some existing methods for this task rely on prior knowledge of individual objects, i.e., the model must be trained on the exact object instance or object category. Others perform unseen object pose estimation but are limited in their feature learning and pose refinement ability. To address these problems, we propose an unseen object pose estimation method that follows a coarse-to-fine framework and leverages the powerful learning ability of diffusion models. We introduce a diffusion model for generating object poses, and conduct a comparison between the generated poses and the original pose to determine the optimal one. We design a novel pose estimation module to provide coarse poses for the PoseDiffusion. This module comprises two feature extraction modules that extract global and masked features. In addition, we propose a strategy to estimate the pose by comparing the similarity between rendered and query poses. The renderings of an unseen object from various viewpoints are generated from its computer-aided design (CAD) model. Our method requires a CAD model of the unseen object only during inference, a scenario well suited to industrial applications. Experimental evaluation on benchmark datasets demonstrates that the proposed framework outperforms existing approaches, achieving state-of-the-art performance in six-DoF object pose estimation.
- 28 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73 (00189456) , 1-14
Joining Spatial Deformable Convolution and a Dense Feature Pyramid for Surface Defect Detection
Although deep learning-based surface defect detection approaches have performed remarkably well in recent years, the complicated shapes and large size differences of surface defects still pose enormous challenges for most existing methods. To address these issues, we propose a novel surface defect detection method joining spatial deformable convolution and a dense feature pyramid, named SDDF-Net. First, we construct a spatial deformable convolution-based feature extraction network, which uses a dynamic convolutional kernel with spatial information to increase the feature extraction capability of complicated defects. Second, we build a dense feature pyramid-based feature fusion network that fuses features from different network layers to improve the detection accuracy of multiscale defects. Third, we present a novel hybrid loss that combines complete intersection over union loss and normalized Wasserstein distance loss to enhance the defect recognition and location learning abilities of our method. Finally, we run our method on the NEU-DET, DAGM2007 and DeepPCB datasets to conduct a comprehensive comparison with some state-of-the-art general object detection models and specialized surface defect detection methods. The experimental results show that the proposed SDDF-Net performs competitively in terms of detection accuracy and computational efficiency when compared with existing methods. This indicates that SDDF-Net achieves good results for surface defect detection and is qualified for real-time processing in industry.
- 28 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 73 (00189456) , 1-14
Joining Spatial Deformable Convolution and a Dense Feature Pyramid for Surface Defect Detection
Although deep learning-based surface defect detection approaches have performed remarkably well in recent years, the complicated shapes and large size differences of surface defects still pose enormous challenges for most existing methods. To address these issues, we propose a novel surface defect detection method joining spatial deformable convolution and a dense feature pyramid, named SDDF-Net. First, we construct a spatial deformable convolution-based feature extraction network, which uses a dynamic convolutional kernel with spatial information to increase the feature extraction capability of complicated defects. Second, we build a dense feature pyramid-based feature fusion network that fuses features from different network layers to improve the detection accuracy of multiscale defects. Third, we present a novel hybrid loss that combines complete intersection over union loss and normalized Wasserstein distance loss to enhance the defect recognition and location learning abilities of our method. Finally, we run our method on the NEU-DET, DAGM2007 and DeepPCB datasets to conduct a comprehensive comparison with some state-of-the-art general object detection models and specialized surface defect detection methods. The experimental results show that the proposed SDDF-Net performs competitively in terms of detection accuracy and computational efficiency when compared with existing methods. This indicates that SDDF-Net achieves good results for surface defect detection and is qualified for real-time processing in industry.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 4559-4564
An efficient grasp pose detection method based on Point-PN
Grasp pose detection is crucial in robotics and automation, but generating reliable and humanoid grasps in unstructured environments remains a challenge. Deep learning neural networks have gained attention in computer vision, leading to the development of grasp detection methods. Existing methods generate multiple potential grasp poses, but in practical applications, only one suitable grasp position is typically chosen, resulting in wasted resources and decreased effectiveness. Moreover, these methods often suffer from computational overhead. In this paper, we propose an efficient grasp pose detection method based on Point-PN to solve these problems. We are able to generate a single optimal grasp pose for objects in the scene by decoupling the grasp pose prediction task into subtasks of grasp approach vector, grasp rotation angle, and grasp depth, and evaluating the pose quality using grasp confidence scores. Additionally, We introduce a multi-structure network that utilizes Point-PN as the network backbone, enabling efficient learning of point cloud features and reducing training costs and resource consumption. For our network, we proposed a joint loss function in order to better adapt to the various grasp pose prediction subtasks. Experimental results demonstrate significant achievements in network evaluation and real robot experiments.
- 17 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 4559-4564
An efficient grasp pose detection method based on Point-PN
Grasp pose detection is crucial in robotics and automation, but generating reliable and humanoid grasps in unstructured environments remains a challenge. Deep learning neural networks have gained attention in computer vision, leading to the development of grasp detection methods. Existing methods generate multiple potential grasp poses, but in practical applications, only one suitable grasp position is typically chosen, resulting in wasted resources and decreased effectiveness. Moreover, these methods often suffer from computational overhead. In this paper, we propose an efficient grasp pose detection method based on Point-PN to solve these problems. We are able to generate a single optimal grasp pose for objects in the scene by decoupling the grasp pose prediction task into subtasks of grasp approach vector, grasp rotation angle, and grasp depth, and evaluating the pose quality using grasp confidence scores. Additionally, We introduce a multi-structure network that utilizes Point-PN as the network backbone, enabling efficient learning of point cloud features and reducing training costs and resource consumption. For our network, we proposed a joint loss function in order to better adapt to the various grasp pose prediction subtasks. Experimental results demonstrate significant achievements in network evaluation and real robot experiments.