Zhongze Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhongze Wang has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Lujian Yao
East China University of Science and Technology, Shanghai, China
11 shared publications - Haitao Zhao
School of Information Science and Engineering, East China University of Science and Technology, Shanghai, China
10 shared publications - Jingchao Peng
School of Information Science and Engineering, East China University of Science and Technology, Shanghai, China
9 shared publications - Kaijie Zhao
East China University of Science and Technology, Shanghai, China
8 shared publications Jingchao Peng
3 shared publicationsHaitao Zhao
2 shared publications
Latest Publications
- 3 November 2025
- Published by Elsevier in Knowledge-Based Systems
- Vol. 331
Bridging element fragmentation and inter-view discontinuity via directional geometric embeddings for cross-modal map construction
- 3 November 2025
- Published by Elsevier in Knowledge-Based Systems
- Vol. 331
Bridging element fragmentation and inter-view discontinuity via directional geometric embeddings for cross-modal map construction
- 20 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48 (2) , 1842-1859
Orthogonal Decoupling Contrastive Regularization: Toward Uncorrelated Feature Decoupling for Unpaired Image Restoration
Unpaired image restoration (UIR) is a significant task due to the difficulty of acquiring paired degraded/clear images with identical backgrounds. In this paper, we propose a novel UIR method based on the assumption that an image contains both degradation-related features, which affect the level of degradation, and degradation-unrelated features, such as texture and semantic information. Our method aims to ensure that the degradation-related features of the restoration result closely resemble those of the clear image, while the degradation-unrelated features align with the input degraded image. Specifically, we introduce a Feature Orthogonalization Module optimized on Stiefel manifold to decouple image features, ensuring feature uncorrelation. A task-driven Depth-wise Feature Classifier is proposed to assign weights to uncorrelated features based on their relevance to degradation prediction. To avoid the dependence of the training process on the quality of the clear image in a single pair of input data, we propose to maintain several degradation-related proxies describing the degradation level of clear images to enhance the model’s robustness. Finally, a weighted PatchNCE loss is introduced to pull degradation-related features in the output image toward those of clear images, while bringing degradation-unrelated features close to those of the degraded input.
- 20 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48 (2) , 1842-1859
Orthogonal Decoupling Contrastive Regularization: Toward Uncorrelated Feature Decoupling for Unpaired Image Restoration
Unpaired image restoration (UIR) is a significant task due to the difficulty of acquiring paired degraded/clear images with identical backgrounds. In this paper, we propose a novel UIR method based on the assumption that an image contains both degradation-related features, which affect the level of degradation, and degradation-unrelated features, such as texture and semantic information. Our method aims to ensure that the degradation-related features of the restoration result closely resemble those of the clear image, while the degradation-unrelated features align with the input degraded image. Specifically, we introduce a Feature Orthogonalization Module optimized on Stiefel manifold to decouple image features, ensuring feature uncorrelation. A task-driven Depth-wise Feature Classifier is proposed to assign weights to uncorrelated features based on their relevance to degradation prediction. To avoid the dependence of the training process on the quality of the clear image in a single pair of input data, we propose to maintain several degradation-related proxies describing the degradation level of clear images to enhance the model’s robustness. Finally, a weighted PatchNCE loss is introduced to pull degradation-related features in the output image toward those of clear images, while bringing degradation-unrelated features close to those of the degraded input.
- 17 October 2025
- Published by Elsevier in Pattern Recognition
- Vol. 172
Prototype-based scatter learning for smoke segmentation
- 17 October 2025
- Published by Elsevier in Pattern Recognition
- Vol. 172
Prototype-based scatter learning for smoke segmentation
- 15 April 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 281
PBMA: Enhancing 3D point cloud tracking with Point-to-Box Motion Augmentation
- 15 April 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 281
PBMA: Enhancing 3D point cloud tracking with Point-to-Box Motion Augmentation
- 7 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-11
Enhancing Point Cloud Tracking With Spatio-Temporal Triangle Optimization
Single Object Tracking (SOT) is a critical task in computer vision, playing a substantial role in fields such as autonomous driving. It represents a spatio-temporal learning challenge, aimed at tracking a designated target, indicated by a bounding box (BBox), in video sequences of 3D point cloud data. Recent developments in the target-motion-estimation (TME) paradigm for SOT have achieved significant performance improvements. This paradigm involves predicting the relative target motion (RTM), which captures the shifts in the target’s center and orientation between consecutive frames. However, existing TME-based methods utilize separate and complex optimizations to regress the RTM, with the RTM’s coordinates shift and rotation angle representing motion properties. Furthermore, RTM fails to offer a comprehensive representation of target motion, such as the motion patterns of objects with varying sizes and movement ranges. Additionally, the current paradigm for multiframe tracking is sensitive to noise and distractors, due to the unrobust proposal selection method based on point counting. To address these limitations, a spatio-temporal triangle optimization method is proposed. All optimization steps from existing TME-based methods are integrated into a spatio-temporal triangle optimization, simplifying the process and improving integration. Diagonals (BBox corners) are used to denote the BBox parameters, allowing the RTM to be represented as a spatiotemporal area swept by diagonals, thus providing a consistent motion measurement and comprehensive motion patterns. A novel proposal selection method is introduced, selecting proposals based on the highest Intersection over Union (IoU) with the ground truth (GT), ensuring more robust tracking in scenarios with multiple distractors. Extensive experiments demonstrate that the proposed spatio-temporal triangle optimization method significantly enhances tracking performance, resulting in improvements of ↑1.5/↑2.2 on the KITTI dataset and ↑2.22/↑2.45 on nuScenes for the two-frame model, and ↑2.8/↑2.4 on KITTI and ↑3.46/↑4.47 on nuScenes for the multi-frame model, achieving state-of-the-art results on both datasets.
- 7 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-11
Enhancing Point Cloud Tracking With Spatio-Temporal Triangle Optimization
Single Object Tracking (SOT) is a critical task in computer vision, playing a substantial role in fields such as autonomous driving. It represents a spatio-temporal learning challenge, aimed at tracking a designated target, indicated by a bounding box (BBox), in video sequences of 3D point cloud data. Recent developments in the target-motion-estimation (TME) paradigm for SOT have achieved significant performance improvements. This paradigm involves predicting the relative target motion (RTM), which captures the shifts in the target’s center and orientation between consecutive frames. However, existing TME-based methods utilize separate and complex optimizations to regress the RTM, with the RTM’s coordinates shift and rotation angle representing motion properties. Furthermore, RTM fails to offer a comprehensive representation of target motion, such as the motion patterns of objects with varying sizes and movement ranges. Additionally, the current paradigm for multiframe tracking is sensitive to noise and distractors, due to the unrobust proposal selection method based on point counting. To address these limitations, a spatio-temporal triangle optimization method is proposed. All optimization steps from existing TME-based methods are integrated into a spatio-temporal triangle optimization, simplifying the process and improving integration. Diagonals (BBox corners) are used to denote the BBox parameters, allowing the RTM to be represented as a spatiotemporal area swept by diagonals, thus providing a consistent motion measurement and comprehensive motion patterns. A novel proposal selection method is introduced, selecting proposals based on the highest Intersection over Union (IoU) with the ground truth (GT), ensuring more robust tracking in scenarios with multiple distractors. Extensive experiments demonstrate that the proposed spatio-temporal triangle optimization method significantly enhances tracking performance, resulting in improvements of ↑1.5/↑2.2 on the KITTI dataset and ↑2.22/↑2.45 on nuScenes for the two-frame model, and ↑2.8/↑2.4 on KITTI and ↑3.46/↑4.47 on nuScenes for the multi-frame model, achieving state-of-the-art results on both datasets.
- 30 September 2024
- Published by Springer Nature
- p. 467-484
DSA: Discriminative Scatter Analysis for Early Smoke Segmentation
- 30 September 2024
- Published by Springer Nature
- p. 467-484
DSA: Discriminative Scatter Analysis for Early Smoke Segmentation
- Open Access
ODCR: Orthogonal Decoupling Contrastive Regularization for Unpaired Image Dehazing
Unpaired image dehazing (UID) holds significant research importance due to the challenges in acquiring haze/clear image pairs with identical backgrounds. This paper proposes a novel method for UID named Orthogonal Decoupling Contrastive Regularization (ODCR). Our method is grounded in the assumption that an image consists of both haze-related features, which influence the degree of haze, and haze-unrelated features, such as texture and semantic information. ODCR aims to ensure that the haze-related features of the dehazing result closely resemble those of the clear image, while the haze-unrelated features align with the input hazy image. To accomplish the motivation, Orthogonal MLPs optimized geometrically on the Stiefel manifold are proposed, which can project image features into an orthogonal space, thereby reducing the relevance between different features. Furthermore, a task-driven Depth-wise Feature Classifier (DWFC) is proposed, which assigns weights to the orthogonal features based on the contribution of each channel's feature in predicting whether the feature source is hazy or clear in a self-supervised fashion. Finally, a Weighted PatchNCE (WPNCE) loss is introduced to achieve the pulling of haze-related features in the output image toward those of clear images, while bringing haze-unrelated features close to those of the hazy input. Extensive experiments demonstrate the superior performance of our ODCR method on UID.
Open AccessODCR: Orthogonal Decoupling Contrastive Regularization for Unpaired Image Dehazing
Unpaired image dehazing (UID) holds significant research importance due to the challenges in acquiring haze/clear image pairs with identical backgrounds. This paper proposes a novel method for UID named Orthogonal Decoupling Contrastive Regularization (ODCR). Our method is grounded in the assumption that an image consists of both haze-related features, which influence the degree of haze, and haze-unrelated features, such as texture and semantic information. ODCR aims to ensure that the haze-related features of the dehazing result closely resemble those of the clear image, while the haze-unrelated features align with the input hazy image. To accomplish the motivation, Orthogonal MLPs optimized geometrically on the Stiefel manifold are proposed, which can project image features into an orthogonal space, thereby reducing the relevance between different features. Furthermore, a task-driven Depth-wise Feature Classifier (DWFC) is proposed, which assigns weights to the orthogonal features based on the contribution of each channel's feature in predicting whether the feature source is hazy or clear in a self-supervised fashion. Finally, a Weighted PatchNCE (WPNCE) loss is introduced to achieve the pulling of haze-related features in the output image toward those of clear images, while bringing haze-unrelated features close to those of the hazy input. Extensive experiments demonstrate the superior performance of our ODCR method on UID.
- Open Access
- 24 March 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 38 (7) , 6621-6629
FoSp: Focus and Separation Network for Early Smoke Segmentation
Early smoke segmentation (ESS) enables the accurate identification of smoke sources, facilitating the prompt extinguishing of fires and preventing large-scale gas leaks. But ESS poses greater challenges than conventional object and regular smoke segmentation due to its small scale and transparent appearance, which can result in high miss detection rate and low precision. To address these issues, a Focus and Separation Network (FoSp) is proposed. We first introduce a Focus module employing bidirectional cascade which guides low-resolution and high-resolution features towards mid-resolution to locate and determine the scope of smoke, reducing the miss detection rate. Next, we propose a Separation module that separates smoke images into a pure smoke foreground and a smoke-free background, enhancing the contrast between smoke and background fundamentally, improving segmentation precision. Finally, a Domain Fusion module is developed to integrate the distinctive features of the two modules which can balance recall and precision to achieve high F_beta. Futhermore, to promote the development of ESS, we introduce a high-quality real-world dataset called SmokeSeg, which contains more small and transparent smoke than the existing datasets. Experimental results show that our model achieves the best performance on three available smoke segmentation datasets: SYN70K (mIoU: 83.00%), SMOKE5K (F_beta: 81.6%) and SmokeSeg (F_beta: 72.05%). The code can be found at https://github.com/LujianYao/FoSp.
- 24 March 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 38 (7) , 6621-6629
Open AccessFoSp: Focus and Separation Network for Early Smoke Segmentation
Early smoke segmentation (ESS) enables the accurate identification of smoke sources, facilitating the prompt extinguishing of fires and preventing large-scale gas leaks. But ESS poses greater challenges than conventional object and regular smoke segmentation due to its small scale and transparent appearance, which can result in high miss detection rate and low precision. To address these issues, a Focus and Separation Network (FoSp) is proposed. We first introduce a Focus module employing bidirectional cascade which guides low-resolution and high-resolution features towards mid-resolution to locate and determine the scope of smoke, reducing the miss detection rate. Next, we propose a Separation module that separates smoke images into a pure smoke foreground and a smoke-free background, enhancing the contrast between smoke and background fundamentally, improving segmentation precision. Finally, a Domain Fusion module is developed to integrate the distinctive features of the two modules which can balance recall and precision to achieve high F_beta. Futhermore, to promote the development of ESS, we introduce a high-quality real-world dataset called SmokeSeg, which contains more small and transparent smoke than the existing datasets. Experimental results show that our model achieves the best performance on three available smoke segmentation datasets: SYN70K (mIoU: 83.00%), SMOKE5K (F_beta: 81.6%) and SmokeSeg (F_beta: 72.05%). The code can be found at https://github.com/LujianYao/FoSp.
- 26 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 26 (15209210) , 7673-7686
DFR-Net: Density Feature Refinement Network for Image Dehazing Utilizing Haze Density Difference
In the image dehazing task, the haze density is a key feature that affects the performance of dehazing methods. The haze density difference, which has rarely been utilized in previous methods, can guide networks to perceive different global densities and focus on local areas with high density or that are difficult to dehaze. In this paper, we propose a density-aware dehazing method named the Density Feature Refinement Network (DFR-Net), which extracts haze density features from density differences and leverages density differences to refine density features. In DFR-Net, we first generate a proposal image that has a lower overall density than the hazy input, resulting in global density differences. Additionally, the dehazing residual of the proposal image reflects the level of dehazing performance and provides local density differences that indicate localized hard dehazing or high-density areas. Subsequently, we introduce a Global Branch (GB) and a Local Branch (LB) to achieve density awareness. In GB, we use Siamese networks for feature extraction of hazy inputs and proposal images, and we propose a Global Density Feature Refinement (GDFR) module that can refine features by pushing features with different global densities further away. In LB, we explore local density features from the dehazing residuals between hazy inputs and proposal images and introduce an Intermediate Dehazing Residual Feedforward (IDRF) module to update local features and pull them close to clear image features. Sufficient experiments demonstrate that the proposed method outperforms state-of-the-art methods on various datasets.
- 26 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 26 (15209210) , 7673-7686
DFR-Net: Density Feature Refinement Network for Image Dehazing Utilizing Haze Density Difference
In the image dehazing task, the haze density is a key feature that affects the performance of dehazing methods. The haze density difference, which has rarely been utilized in previous methods, can guide networks to perceive different global densities and focus on local areas with high density or that are difficult to dehaze. In this paper, we propose a density-aware dehazing method named the Density Feature Refinement Network (DFR-Net), which extracts haze density features from density differences and leverages density differences to refine density features. In DFR-Net, we first generate a proposal image that has a lower overall density than the hazy input, resulting in global density differences. Additionally, the dehazing residual of the proposal image reflects the level of dehazing performance and provides local density differences that indicate localized hard dehazing or high-density areas. Subsequently, we introduce a Global Branch (GB) and a Local Branch (LB) to achieve density awareness. In GB, we use Siamese networks for feature extraction of hazy inputs and proposal images, and we propose a Global Density Feature Refinement (GDFR) module that can refine features by pushing features with different global densities further away. In LB, we explore local density features from the dehazing residuals between hazy inputs and proposal images and introduce an Intermediate Dehazing Residual Feedforward (IDRF) module to update local features and pull them close to clear image features. Sufficient experiments demonstrate that the proposed method outperforms state-of-the-art methods on various datasets.