Tube Defect Detection Algorithm Under Noisy Environment Using Feature Vector and Neural Networks
- 10 Downloads
Surface flaw detection has been advanced steadily for decades thank to the advent of computer vision and artificial intelligence. However, there exist serious defect detection challenges in tube manufacturing, including the lack of a collected dataset, decision-making ambiguity in engineering judgment, and unstable lighting condition of the environment. This work aims to investigate an effective method to distinguish deformity that performs despite these challenges to deliver quality control in tube manufacturing. We present a new tube detection algorithm under limited data set and noisy environment due to unstable lighting condition, for which we introduced a feature vector to describe the defect problem. Using the feature vector and a neural network we are able to successfully detect and classify tube defect.
KeywordsMachine vision Surface flaw detection Neural network
This work was supported by the 2016 Research Fund of University of Ulsan.
- 3.Erhan, D., Szegedy, C., Toshev, A., & Anguelov, D. (2013). Scalable object detection using deep neural networks. arXiv:1312.2249.
- 5.Luiz, A. O. M., Flavio, L. C. P., & Paulo, E. M. A. (2010). Automatic detection of surface defects on rolled steel using computer vision and artificial neural networks. In IEEE (pp. 1081–1086). https://doi.org/10.1109/iecon.2010.5675519.
- 11.A practical guide to machine vision lighting: National instruments. http://www.ni.com/white-paper/6901/en/. Accessed May 5, 2018.
- 12.Lee, D. J., Redd, S., Schoenberger, R., Xu, X., & Zhan, P. (2003). An automated fish species classification and migration monitoring system. In The 29th annual conference of the IEEE industrial electronics society, 2003. IECON’03 (Vol. 2, pp. 1080–1085).Google Scholar
- 13.Yang, S., Zhang, C., & Wu, W. (2018). Binary output layer of feedforward neural networks for solving multi-class classification problems. arXiv:1801.07599[cs, math].