Recently, the top academic journal in the field of artificial intelligence—Engineering Applications of Artificial Intelligence online published the latest research results of Professor Yunhui Yan's research group (Machine Vision and Robotics Laboratory) from the School of Mechanical Engineering and Automation, NEU in the field of intelligent detection of industrial surface crack defects. The research results are entitled EAFNet: Extraction-Amplification-Fusion Network for Tiny Cracks Detection. Ziang Zhou, an undergraduate student of Grade 2020 from the School of Mechanical Engineering and Automation, is the first author, Wensong Zhao, an undergraduate student of Grade 2020 from such School, is the second author, and Associate Professor Kechen Song and Teacher Jun Li are the tutors.
Intelligent detection technology of tiny crack defects plays important role in the field of industrial vision detection. However, there is a lack of image datasets and detection methods specifically for tiny crack defects. In this research, the image of tiny cracks in industrial applications was defined, and a dataset of images of tiny crack defects was constructed. This dataset contains many images of tiny crack defects under different interference backgrounds, and it is the first image data set dedicated to tiny crack defects at present. In addition, a three-stage detection method of extraction-amplification-fusion was proposed, and it can effectively detect tiny cracks.
Additionally, this research group of the undergraduate students published a paper entitled SPCNet: A Strip Pyramid ConvNeXt Network for Detection of Road Surface Defects on the journal—Signal Image and Video Processing, and won the "Excellent Academic Paper" award in the 7th Academic Paper Competition of the NEU in 2024. The above research was supported by the College Student Innovation and Entrepreneurship Training Program Project (project name: Research on Expressway Defect Detection Algorithm based on Deep Learning, with project number: 221180).
