Kaige Zhang

Dr. Kaige Zhang


Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

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

Representative Publications

View All Publications
1 March 2018 in Journal of Computing in Civil Engineering
Verified
Journal Article
Closed access
KZKaige Zhang
Kaige Zhang
MXMin Xian
Min Xian
BZBoyu Zhang
Boyu Zhang
This work focuses on solving two challenging problems in pavement crack detection: (1) noises caused by complicated pavement textures and intensity inhomogeneity cannot be removed effectively, which makes crack extraction difficult; and (2) sealed cracks and cracks with similar intensity and width cannot be separated correctly, which makes data analysis and budgeting inaccurate. Here, a unified crack and sealed crack detection approach is proposed that can detect and separate both cracks and sealed cracks under the same framework. It trains a deep convolutional neural network to preclassify a pavement image into crack, sealed crack, and background regions. A blockwise thresholding method is developed to segment the crack/sealed crack pixels efficiently and effectively. Finally, tensor voting–based curve detection is applied to extract the crack/sealed crack. The proposed approach is validated using 800 images (each 2,000×4,000  pixels ); the experimental results demonstrate that this approach accurately distinguishes cracks from sealed cracks and achieves very good detection performance (recall=0.951 ; precision=0.847 ).

7 May 2020 in IEEE Transactions on Intelligent Transportation Systems
Verified
Journal Article
Closed access
KZKaige Zhang
Kaige Zhang
YZYingtao Zhang
Yingtao Zhang
MXMin Xian
Min Xian
Fully convolutional network is a powerful tool for per-pixel semantic segmentation/detection. However, it is problematic when coping with crack detection using partially accurate ground truths (GTs): the network may easily converge to the status that treats all the pixels as background (BG) and still achieves a very good loss, named ``All Black'' phenomenon, due to the unavailability of accurate GTs and the data imbalance. To tackle this problem, we propose crack-patch-only (CPO) supervised generative adversarial learning for end-to-end training, which forces the network to always produce crack-GT images while reserves both crack and BG-image translation abilities by feeding a larger-size crack image into an asymmetric U-shape generator to overcome the ``All Black'' issue. The proposed approach is validated using four crack datasets; and achieves state-of-the-art performance comparing with that of the recently published works in efficiency and accuracy.

30 December 2017
Verified
Book Chapter
Closed access
PDF file
KZKaige Zhang
Kaige Zhang
HCHengda Cheng
Hengda Cheng


27 February 2023
Verified
Book Chapter
Closed access
KZKaige Zhang
Kaige Zhang
MXMin Xian
Min Xian
The fully-convolutional network was introduced for pixel-level crack detection; however, the method needs pixel-level annotations which are difficult to obtain for crack-like objects. The commonly used Ground Truths (GTs) are the 1-pixel crack curves that cannot match the real cracks at pixel level, leading to the training to always converge to the status that treats all the pixels as background, named ‘All Black’ issue due to the unavailability of accurate GTs and data imbalance. This chapter proposes Crack-Patch-Only (CPO) supervised generative adversarial learning for end-to-end training, which forces the network to always produce crack-GT images while reserving both crack and BG-image translation abilities by feeding a larger-size crack image into an asymmetric U-shape generator to overcome the ‘All Black’ issue. The chapter provides a concrete example, utilizing generative adversarial learning to resolve practical problem, which is also the first work to perform object regularization by providing a loss that is from data training instead of hand-made regularization term. It is also an important work that could help readers understand generative adversarial learning intuitively. The method used to address the problem can be generalized to other computer vision problems and help in addressing the data imbalance in a data-driven manner which is a very important direction for the community.

27 February 2023
Verified
Book Chapter
Closed access
KZKaige Zhang
Kaige Zhang
MXMin Xian
Min Xian
Cracks are common defects on surfaces of man-made structures, such as pavements, bridges, walls of nuclear power plants, ceilings of tunnels, etc. Timely discovering and repairing of the cracks are of great significance and importance for keeping healthy infrastructures and preventing further damages. Traditionally, the cracking inspection was conducted manually and it was labor-intensive, time-consuming and costly. With the development of Artificial Intelligence (AI), deep learning has achieved great success and has been viewed as the most promising way for crack detection. This book discusses crack detection with deep learning based on the authors’ years of experience on pavement surface inspection. A complete automatic crack detection solution is presented and discussed from scratch which only provides an important reference for the relevant engineers, but also provides an easy-to-go tutorial for beginners to get familiar with deep learning and computer vision.