WCSE

A Fast Workpiece Surface Defect Detection Approach Based on Multiple Hierarchical Features

Abstract—Workpiece surface defect detection is of great importance to the product quality. In this paper, a lightweight convolutional neural network is improved. Based on the multiple hierarchical features extracted by the network, the images are divided into girds with different sizes, which are responsible for the detection of different defects. Furthermore, a new post-processing method is introduced to merge these grids. The experimental results show that our approach achieves good detection results and has a great advantage in inference speed and computing power demand.

Index Terms—defect detection, CNN, anchor selection, workpiece surface defects

Yuxuan Han
School of Data Science Engineering, East China Normal University, Shanghai, CHINA
Jun Guo
Department of Vision Algorithm, Changzhou Micro-Intelligence Co. Ltd., Changzhou, CHINA
Youguang Chen
School of Data Science Engineering, East China Normal University, Shanghai, CHINA

[Download]

Cite:Yuxuan Han, Jun Guo, Youguang Chen, "A Fast Workpiece Surface Defect Detection Approach Based on Multiple Hierarchical Features, " Proceedings of 2022 the 12th International Workshop on Computer Science and Engineering (WCSE 2022), pp. 236-241, June 24-27, 2022.