Ziyang Wang

Dr. Ziyang Wang

orcid

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 March 2021 in IEEE Access
Verified
Journal Article
Open access
DWDan Wei
Dan Wei
XHXiaoqiang Hu
Xiaoqiang Hu
ZWZiyang Wang
Ziyang Wang
JSJianglin Shen
Jianglin Shen
HRHongjuan Ren
Hongjuan Ren
How to extract discriminative features from redundant video information is a key issue for video pedestrian re-identification. Factors such as occlusion, perspective, and posture changes in complex environments pose severe challenges to pedestrian re-identification based on local methods. In this paper, a posture-guided multi-scale structural relationship learning pedestrian re-identification method is proposed. The purpose is to analyze the video sequence of pedestrians based on the reference pose and the pose alignment model, and extract the sample frame with the highest image quality and the most complete spatial information in the reference pose. The method based on posture guidance can more accurately eliminate the interference of background, occlusion and perspective factors. To further explore the potential relationship between local regions, this paper calculates the relationship matrix between the local regions based on the relationship model to further calculate the relationship weight, and the graph convolutional network based on the relationship weight learns the structural relationship feature of multi-scale regions. The input of the graph convolutional network is a local region divided by a multi-scale method, and the output is a pose-guided multi-scale structural relationship feature. The experimental results on three public datasets show that the proposed method performs favorably against state-of-the-art methods.

7 September 2026 in Algorithms
Verified
Journal Article
Open access
PDF file
ZWZiyang Wang
Ziyang Wang
HWHaoze Wu
Haoze Wu
DWDan Wei
Dan Wei
Existing video person re-identification (ReID) methods still face three major challenges: limited spatial perception from RGB-only inputs, equal-weight local feature extraction under occlusion, and ineffective fusion of heterogeneous motion and appearance features. To address these issues, we propose an RGB-D Triple Pyramid Model (TPM) for robust spatiotemporal feature learning. The proposed framework can utilize conventional RGB surveillance videos without requiring additional depth sensors. Specifically, TPM first generates pseudo RGB-D representations from RGB frames through monocular depth estimation. This process introduces additional spatial cues, such as pedestrian scale and relative distance, to improve spatial perception. To capture temporal dynamics, frame differences from consecutive 5-frame clips are extracted and aggregated into compact motion representations, which effectively describe short-term pedestrian movements. For appearance representation, each pedestrian frame is divided into dense uniform grids to extract fine-grained local features. The proposed method automatically identifies visible and informative regions and assigns adaptive weights to different grid cells. These weighted local features are aggregated to construct an occlusion-robust appearance representation, which emphasizes discriminative regions while reducing the influence of unreliable areas. Furthermore, a Transferable Feature Fusion (TFF) module is introduced to integrate motion and appearance representations. Instead of directly combining heterogeneous features, TFF transfers discriminative information from motion features to appearance features by exploiting the regression residuals learned from the motion branch. This process refines appearance representations and improves the complementary interaction between motion and appearance features. Extensive experiments on three widely used video ReID benchmarks, including MARS, iLIDS-VID, and PRID-2011, demonstrate the effectiveness of the proposed method. Specifically, TPM achieves improvements of 4.7%, 4.8%, and 1.7% in Rank-1, Rank-5, and Rank-20 accuracy, respectively, on the MARS dataset, validating the effectiveness of the proposed framework.