Since workpiece defect detection is a typical problem in computer vision with small datasets, generally its solutions cannot exploit the advantages of high accuracy, generalization ability, and neural network structures from the deep learning paradigm. Thus, traditional image processing techniques are still widely applied in such requirements. Aiming at three types of defects (crack, pitting and scratch) on a workpiece with non-concentric circular textures that severely interfere in the defect recognition stage, this paper proposes a sliding window filter for the texture detection. Experiments compare the proposed method with the polar coordinate mapping method and the T-smooth texture removal algorithm. Results show that the proposed method reveals the three types of defects better than the other two methods.
CITATION STYLE
Qin, S., Guo, D., Chen, H., & Xi, N. (2019). Non-concentric Circular Texture Removal for Workpiece Defect Detection. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11743 LNAI, pp. 576–584). Springer Verlag. https://doi.org/10.1007/978-3-030-27538-9_49
Mendeley helps you to discover research relevant for your work.