Wenxuan Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Wenxuan Wang has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Chunhui Zhao
College of Information and Communication Engineering, Harbin Engineering University, Harbin, China
5 shared publications - Yiming Yan
Harbin Engineering University
5 shared publications - Wei Hou
System Demonstration Center, Harbin Space Star Data System Technology Company Ltd., Harbin, China
4 shared publications - Nan Su
Harbin Engineering University
2 shared publications - Qingyu Xia
Key Laboratory of Advanced Marine Communication and Information Technology, Ministry of Industry and Information Technology, and College of Information and Communication Engineering, Harbin Engineering University, Harbin, China
2 shared publications - Baoyu Ge
Key Laboratory of Advanced Marine Communication and Information Technology, Ministry of Industry and Information, College of Information and Communication Engineering, Harbin Engineering University, Harbin 150009, China
1 shared publication
Latest Publications
- 4 July 2026
- Published by Springer Nature
- p. 155-165
DSPLR: An Novel Unsupervised Visible-Infrared Person Re-identification Framework for Aerial Image
- 4 July 2026
- Published by Springer Nature
- p. 155-165
DSPLR: An Novel Unsupervised Visible-Infrared Person Re-identification Framework for Aerial Image
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 15667-15678
Dual-Granularity Consistency Optimization Framework for Unsupervised Aerial Cross-Modality Person Re-Identification
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 15667-15678
Dual-Granularity Consistency Optimization Framework for Unsupervised Aerial Cross-Modality Person Re-Identification
- Open Access
- 5 June 2025
- Published by MDPI AG in Remote Sensing
- Vol. 17 (11) , 1956
A Re-Identification Framework for Visible and Thermal-Infrared Aerial Remote Sensing Images with Large Differences of Elevation Angles
Visible and thermal-infrared re-identification (VTI-ReID) based on aerial images is a challenging task due to the large range of elevation angles, which exacerbates the modality differences between different modalities. The substantial modality gap makes it challenging for existing methods to extract identity information from aerial images captured at wide elevation angles. This limitation significantly reduces VTI-ReID accuracy. This issue is particularly pronounced in elongated targets. To address this issue, a robust framework for extracting identity representation (RIRE) is proposed, specifically designed for VTI-ReID in aerial cross-modality images. This framework adopts a mapping method based on global representation decomposition and local representation aggregation. It effectively extracts features related to identity from aerial images and aligns the global representations of images captured from different angles within the same identity space. This approach enhances the adaptability of the VTI-ReID task to elevation angle differences. To validate the effectiveness of the proposed framework, a dataset group for elongated target VTI-ReID based on unmanned aerial vehicle (UAV)-captured data has been created. Extensive evaluations of the proposed framework on the proposed dataset group indicate that the framework significantly improves the robustness of the extracted identity information for elongated targets in aerial images, thereby enhancing the accuracy of VTI-ReID.
- 5 June 2025
- Published by MDPI AG in Remote Sensing
- Vol. 17 (11) , 1956
Open AccessA Re-Identification Framework for Visible and Thermal-Infrared Aerial Remote Sensing Images with Large Differences of Elevation Angles
Visible and thermal-infrared re-identification (VTI-ReID) based on aerial images is a challenging task due to the large range of elevation angles, which exacerbates the modality differences between different modalities. The substantial modality gap makes it challenging for existing methods to extract identity information from aerial images captured at wide elevation angles. This limitation significantly reduces VTI-ReID accuracy. This issue is particularly pronounced in elongated targets. To address this issue, a robust framework for extracting identity representation (RIRE) is proposed, specifically designed for VTI-ReID in aerial cross-modality images. This framework adopts a mapping method based on global representation decomposition and local representation aggregation. It effectively extracts features related to identity from aerial images and aligns the global representations of images captured from different angles within the same identity space. This approach enhances the adaptability of the VTI-ReID task to elevation angle differences. To validate the effectiveness of the proposed framework, a dataset group for elongated target VTI-ReID based on unmanned aerial vehicle (UAV)-captured data has been created. Extensive evaluations of the proposed framework on the proposed dataset group indicate that the framework significantly improves the robustness of the extracted identity information for elongated targets in aerial images, thereby enhancing the accuracy of VTI-ReID.
- 29 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 31 (10709908) , 1254-1258
A Three-Stage Framework for Video-Based Visible-Infrared Person Re-Identification
Video-based visible-infrared person re-identification (VI-ReID) aims to identify the same suspicious person captured by different sensors at different times and scenes. Overcoming the modality discrepancy caused by the different imaging principles of visible and infrared videos is a crucial challenge in this field. To address the above challenge, we propose a three-stage framework for video-based VI-ReID to fully exploit the identity information contained in different modality data, which named Decomposition-Mining- Aggregation framework. Our framework consists of three stages: decomposition, mining, and aggregation stage. Experimental results demonstrate that the combination of the three stages can effectively enhance the feature extraction network's ability to extract identity information from images with modality differences. In addition, we also illustrate the influence of video frame sampling and exploration strategy on the quality of VI-ReID model through experiments. Extensive experimental results on the HITSZ-VCM dataset demonstrate that our framework achieves the best accuracy in the VI-ReID task.
- 29 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 31 (10709908) , 1254-1258
A Three-Stage Framework for Video-Based Visible-Infrared Person Re-Identification
Video-based visible-infrared person re-identification (VI-ReID) aims to identify the same suspicious person captured by different sensors at different times and scenes. Overcoming the modality discrepancy caused by the different imaging principles of visible and infrared videos is a crucial challenge in this field. To address the above challenge, we propose a three-stage framework for video-based VI-ReID to fully exploit the identity information contained in different modality data, which named Decomposition-Mining- Aggregation framework. Our framework consists of three stages: decomposition, mining, and aggregation stage. Experimental results demonstrate that the combination of the three stages can effectively enhance the feature extraction network's ability to extract identity information from images with modality differences. In addition, we also illustrate the influence of video frame sampling and exploration strategy on the quality of VI-ReID model through experiments. Extensive experimental results on the HITSZ-VCM dataset demonstrate that our framework achieves the best accuracy in the VI-ReID task.
- Open Access
- 31 March 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (7) , 1249
IML-Net: A Framework for Cross-View Geo-Localization with Multi-Domain Remote Sensing Data
Cross-view geolocation is a valuable yet challenging task. In practical applications, the images targeted by cross-view geolocation technology encompass multi-domain remote sensing images, including those from different platforms (e.g., drone cameras and satellites), different perspectives (e.g., nadir and oblique), and different temporal conditions (e.g., various seasons and weather conditions). Based on the characteristics of these images, we have designed an effective framework, Image Reconstruction and Multi-Unit Mutual Learning Net (IML-Net), for accomplishing cross-view geolocation tasks. By incorporating a deconvolutional network into the architecture to reconstruct images, we can better bridge the differences in remote sensing image features across different domains. This enables the mapping of target images from different platforms and perspectives into a shared latent space representation, obtaining more discriminative feature descriptors. The process enhances the robustness of feature extraction for locating targets across a wide range of perspectives. To improve the network’s performance, we introduce attention regions learned from different units as augmented data during the training process. For the current cross-view geolocation datasets, the use of large-scale datasets is limited due to high costs and privacy concerns, leading to the prevalent use of simulated data. However, real data allow the network to learn more generalizable features. To make the model more robust and stable, we collected two groups of multi-domain datasets from the Zurich and Harbin regions, incorporating real data into the cross-view geolocation task to construct the ZHcity750 Dataset. Our framework is evaluated on the cross-domain ZHcity750 Dataset, which shows competitive results compared to state-of-the-art methods.
- 31 March 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (7) , 1249
Open AccessIML-Net: A Framework for Cross-View Geo-Localization with Multi-Domain Remote Sensing Data
Cross-view geolocation is a valuable yet challenging task. In practical applications, the images targeted by cross-view geolocation technology encompass multi-domain remote sensing images, including those from different platforms (e.g., drone cameras and satellites), different perspectives (e.g., nadir and oblique), and different temporal conditions (e.g., various seasons and weather conditions). Based on the characteristics of these images, we have designed an effective framework, Image Reconstruction and Multi-Unit Mutual Learning Net (IML-Net), for accomplishing cross-view geolocation tasks. By incorporating a deconvolutional network into the architecture to reconstruct images, we can better bridge the differences in remote sensing image features across different domains. This enables the mapping of target images from different platforms and perspectives into a shared latent space representation, obtaining more discriminative feature descriptors. The process enhances the robustness of feature extraction for locating targets across a wide range of perspectives. To improve the network’s performance, we introduce attention regions learned from different units as augmented data during the training process. For the current cross-view geolocation datasets, the use of large-scale datasets is limited due to high costs and privacy concerns, leading to the prevalent use of simulated data. However, real data allow the network to learn more generalizable features. To make the model more robust and stable, we collected two groups of multi-domain datasets from the Zurich and Harbin regions, incorporating real data into the cross-view geolocation task to construct the ZHcity750 Dataset. Our framework is evaluated on the cross-domain ZHcity750 Dataset, which shows competitive results compared to state-of-the-art methods.
- Open Access
- 7 June 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (12) , 2974
A Novel Object-Level Building-Matching Method across 2D Images and 3D Point Clouds Based on the Signed Distance Descriptor (SDD)
In this work, a novel object-level building-matching method using cross-dimensional data, including 2D images and 3D point clouds, is proposed. The core of this method is a newly proposed plug-and-play Joint Descriptor Extraction Module (JDEM) that is used to extract descriptors containing buildings’ three-dimensional shape information from object-level remote sensing data of different dimensions for matching. The descriptor is named Signed Distance Descriptor (SDD). Due to differences in the inherent properties of different dimensional data, it is challenging to match buildings’ 2D images and 3D point clouds on the object level. In addition, features extracted from the same building in images taken at different angles are usually not exactly identical, which will also affect the accuracy of cross-dimensional matching. Therefore, the question of how to extract accurate, effective, and robust joint descriptors is key to cross-dimensional matching. Our JDEM maps different dimensions of data to the same 3D descriptor SDD space through the 3D geometric invariance of buildings. In addition, Multi-View Adaptive Loss (MAL), proposed in this paper, aims to improve the adaptability of the image encoder module to images with different angles and enhance the robustness of the joint descriptors. Moreover, a cross-dimensional object-level data set was created to verify the effectiveness of our method. The data set contains multi-angle optical images, point clouds, and the corresponding 3D models of more than 400 buildings. A large number of experimental results show that our object-level cross-dimensional matching method achieves state-of-the-art outcomes.
- 7 June 2023
- Published by MDPI AG in Remote Sensing
- Vol. 15 (12) , 2974
Open AccessA Novel Object-Level Building-Matching Method across 2D Images and 3D Point Clouds Based on the Signed Distance Descriptor (SDD)
In this work, a novel object-level building-matching method using cross-dimensional data, including 2D images and 3D point clouds, is proposed. The core of this method is a newly proposed plug-and-play Joint Descriptor Extraction Module (JDEM) that is used to extract descriptors containing buildings’ three-dimensional shape information from object-level remote sensing data of different dimensions for matching. The descriptor is named Signed Distance Descriptor (SDD). Due to differences in the inherent properties of different dimensional data, it is challenging to match buildings’ 2D images and 3D point clouds on the object level. In addition, features extracted from the same building in images taken at different angles are usually not exactly identical, which will also affect the accuracy of cross-dimensional matching. Therefore, the question of how to extract accurate, effective, and robust joint descriptors is key to cross-dimensional matching. Our JDEM maps different dimensions of data to the same 3D descriptor SDD space through the 3D geometric invariance of buildings. In addition, Multi-View Adaptive Loss (MAL), proposed in this paper, aims to improve the adaptability of the image encoder module to images with different angles and enhance the robustness of the joint descriptors. Moreover, a cross-dimensional object-level data set was created to verify the effectiveness of our method. The data set contains multi-angle optical images, point clouds, and the corresponding 3D models of more than 400 buildings. A large number of experimental results show that our object-level cross-dimensional matching method achieves state-of-the-art outcomes.
- Open Access
- 18 July 2022
- Published by Association for Computing Machinery (ACM)
- p. 202-214
AEON: a method for automatic evaluation of NLP test cases
Due to the labor-intensive nature of manual test oracle construction, various automated testing techniques have been proposed to enhance the reliability of Natural Language Processing (NLP) software. In theory, these techniques mutate an existing test case (e.g., a sentence with its label) and assume the generated one preserves an equivalent or similar semantic meaning and thus, the same label. However, in practice, many of the generated test cases fail to preserve similar semantic meaning and are unnatural (e.g., grammar errors), which leads to a high false alarm rate and unnatural test cases. Our evaluation study finds that 44% of the test cases generated by the state-of-the-art (SOTA) approaches are false alarms. These test cases require extensive manual checking effort, and instead of improving NLP software, they can even degrade NLP software when utilized in model training. To address this problem, we propose AEON for Automatic Evaluation Of NLP test cases. For each generated test case, it outputs scores based on semantic similarity and language naturalness. We employ AEON to evaluate test cases generated by four popular testing techniques on five datasets across three typical NLP tasks. The results show that AEON aligns the best with human judgment. In particular, AEON achieves the best average precision in detecting semantic inconsistent test cases, outperforming the best baseline metric by 10%. In addition, AEON also has the highest average precision of finding unnatural test cases, surpassing the baselines by more than 15%. Moreover, model training with test cases prioritized by AEON leads to models that are more accurate and robust, demonstrating AEON’s potential in improving NLP software.
- 18 July 2022
- Published by Association for Computing Machinery (ACM)
- p. 202-214
Open AccessAEON: a method for automatic evaluation of NLP test cases
Due to the labor-intensive nature of manual test oracle construction, various automated testing techniques have been proposed to enhance the reliability of Natural Language Processing (NLP) software. In theory, these techniques mutate an existing test case (e.g., a sentence with its label) and assume the generated one preserves an equivalent or similar semantic meaning and thus, the same label. However, in practice, many of the generated test cases fail to preserve similar semantic meaning and are unnatural (e.g., grammar errors), which leads to a high false alarm rate and unnatural test cases. Our evaluation study finds that 44% of the test cases generated by the state-of-the-art (SOTA) approaches are false alarms. These test cases require extensive manual checking effort, and instead of improving NLP software, they can even degrade NLP software when utilized in model training. To address this problem, we propose AEON for Automatic Evaluation Of NLP test cases. For each generated test case, it outputs scores based on semantic similarity and language naturalness. We employ AEON to evaluate test cases generated by four popular testing techniques on five datasets across three typical NLP tasks. The results show that AEON aligns the best with human judgment. In particular, AEON achieves the best average precision in detecting semantic inconsistent test cases, outperforming the best baseline metric by 10%. In addition, AEON also has the highest average precision of finding unnatural test cases, surpassing the baselines by more than 15%. Moreover, model training with test cases prioritized by AEON leads to models that are more accurate and robust, demonstrating AEON’s potential in improving NLP software.