Yafei Jian

Yafei Jian

Auto-generated profile, based on publicly available information

Yafei Jian

State Key Laboratory of Safety and Resilience of Civil Engineering in Mountain Area, East China Jiaotong University, Nanchang, 330013, People’s Republic of China

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

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

Representative Publications

View All Publications
9 October 2025 in Engineering Research Express
Verified
Journal Article
Closed access
MCMengcheng Chen
Mengcheng Chen
YJYafei Jian
Yafei Jian
HLHuan Lai
Huan Lai
WFWei Fang
Wei Fang
SHSiyuan Huang
Siyuan Huang
QJQihai Ji
Qihai Ji
Corrosion leads to the performance degradation of metal and its composite structures, such as steel-concrete structures. The manual detection method is susceptible to subjective judgment, and the associated costs are substantial. In order to assess the corrosion evolution during routine maintenance, it is necessary to identify the corrosion zone. In this paper, we propose a framework for identification of a real-world local corrosion zone under HAU-Net, aiming to address the practical challenges associated with accurately measuring the shape and size of such corrosion zones in actual applications. Firstly, to improve the model’s attention to the local corrosion zone, a new segmentation model called HAU-Net Model is developed by adding a self-designed hybrid attention convolution module (HACM) to the U-Net network. Secondly, the image of a real-world local corrosion zone is fused with the LiDAR data of the corrosion zone by using the registration method, obtaining a new image of 1mm pixel scale. Thirdly, the HAU-Net Model is used to test the new image to obtain a binary image. Finally, the shape identification and size calculation of the real-world local corrosion zone are performed based on the binary image. The experimental results show that, the proposed HAU-Net model can fully make use of the channel context information and spatial context information, and obtain a better segmentation performance on the corrosion dataset constructed from available corrosion images; it outperforms the original U-Net model in terms of the accuracy, precision, and MIoU metrics, which is improved by 2.3%, 4.06%, and 3.95%, respectively. Meanwhile, the errors between the predictions and actual measurements of the real-world local corrosion zone in the test are less than 5%, demonstrating the applicability and efficacy of the proposed method.