Zhongheng Wu

Zhongheng Wu

Auto-generated profile, based on publicly available information

Zhongheng Wu

General Motors Canada, Markham, ON, Canada

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

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

Representative Publications

View All Publications
11 March 2026 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
ZZZhe Zeng
Zhe Zeng
SZShuo Zhang
Shuo Zhang
LKLing Ke
Ling Ke
SHSheng Hui
Sheng Hui
DLDan Liu
Dan Liu
QWQing Wei
Qing Wei
Sea fog poses a critical challenge to maritime safety. Current sea fog detection methods based on geostationary satellite data predominantly rely on the visual interpretation of samples, yet conventional visual interpretation methods are prone to subjective biases in sample labeling. This study proposes a novel dual-branch Patch-CNN architecture using Himawari-8 satellite data, which addresses labeling uncertainty through CALIPSO-derived vertical aerosol profiles that unambiguously discriminate sea fog from low-level clouds. Training samples were exclusively constructed from CALIPSO vertical feature masks (VFM), ensuring spectral reliability by eliminating cloud-fog ambiguity. The framework integrates three distinctive features: (1) Spatial-context preservation via patch-based inputs to encapsulate neighborhood characteristics around sample points; (2) multispectral feature extraction through a dual-branch architecture that separately captures characteristics from visible-near infrared and infrared channels; and (3) adaptive feature fusion employing attention-weighted feature integration to enhance discriminative representation. Evaluated across the Yellow-Bohai Sea region, the model achieved 89.74% overall accuracy and 82.98% detection probability, outperforming conventional machine learning and deep learning benchmarks. In validation testing with meteorological station data, the proposed method exhibited stronger robustness than traditional machine learning and deep learning methods, with 87 true positives, while case studies using CALIOP VFM and ICOADS visibility data found that the present network could accurately identify the foggy area. This work establishes a satellite-based fog detection network, particularly suited for complex marine environments where fog-cloud spectral overlap challenges traditional methods.

16 December 2019 in IEEE Access
Verified
Journal Article
Open access
XZXinyan Zou
Xinyan Zou
Zhixiang FangZhixiang Fang
Zhixiang Fang
HZHaoyu Zhong
Haoyu Zhong
ZWZhongheng Wu
Zhongheng Wu
XLXiongyan Liu
Xiongyan Liu
Obtaining trajectory-related data through crowdsourcing is a useful approach for acquiring and updating information on road networks because these data are easily accessible and up to date. However, it is challenging to ensure the accuracy of shape- and connectivity-related information on road networks when this is extracted from massive amounts of data. To address this challenge, this paper proposes an inertia mutation energy model (IMEM) to extract and mine information on road networks based on crowdsourced trajectories by using features of images of vehicular trajectories to mine the connectivity between roads. The central assumption of the model is that a road segment may have inertial energy to extend to another road segment that decreases in case of the change in direction. Three critical steps are involved in extracting information concerning roads. First, images of the trajectory and corresponding feature images are constructed, and high-quality trajectories are filtered to split the road into several segments. Second, the IMEM is proposed to measure the potential extension of each road segment with the aim of mining connectivity-related information between the given trajectories. Finally, the centerline of the road is obtained using mathematical morphology and a thinning algorithm. The proposed algorithm was tested by using Global Position System trajectories of the Didi Taxi in Wuhan, China, and the results show that it reduced time cost by over 99% compared with vector algorithms proposed in the literature. Moreover, it enhanced the precision of the results of extraction by 10%-20% compared with traditional kernel density evaluation algorithms.

14 January 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
ZZZhe Zeng
Zhe Zeng
QWQing Wei
Qing Wei
WTWenxia Tan
Wenxia Tan
YSYongtian Shen
Yongtian Shen
ZWZhongheng Wu
Zhongheng Wu
Wind turbines are vital equipment for converting wind energy into electrical energy. Precise detection of wind turbines in high-resolution images is of great importance for building accurate wind turbine databases. However, their distinctive white appearance poses challenges for detection in snowy domain images. To address this problem, a domain adaptive deep convolutional neural network model named DAWindNet, is proposed, aiming at cross-domain extraction of wind turbines in snow background images. First, in the image-level module, the model learns discrepancies between images from different domains, alters the image style, and achieves data alignment for complete images. Second, in the instance-level module, the model focus shifts to the structural information of wind turbine targets, further refining cross-domain data alignment. Perception loss and domain difference loss are designed to preserve semantic consistency across domains and mitigate domain offset phenomena of the target. Finally, bidirectional feature pyramid network and an attention mechanism are incorporated to enhance the network's ability to extract wind power features and achieve higher recognition rates. Experimental results on datasets representing ordinary, bare soil, and snowy domains validate that DAWindNet achieves satisfactory performance, with a recall rate of 63.8% and an average precision (AP) of 67.3% when transitioning from ordinary to snow background domains, and a recall rate of 65.1% and an AP of 66.1% for the transition between bare soil and snow background domains. These results demonstrate the effectiveness of the proposed modifications for the cross-domain extraction of wind turbines