Zhang Daode
School of Mechanical Engineering, Hubei University of Technology, Wuhan, 430068, Hubei, China
Xue Yangliu
School of Mechanical Engineering, Hubei University of Technology, Wuhan, 430068, Hubei, China
Hu Xinyu
School of Mechanical Engineering, Hubei University of Technology, Wuhan, 430068, Hubei, China
ABSTRACT
Based on cognition and research for vision measurement method on chips defects, a new type of chip defects detection algorithm is proposed. In view of defect characteristic and examination request of chips surface, the first is pre-processed on the images by CMOS camera. This process includes gradation, median blur, iterative segmentation and contours extraction on images, which will get the chips contours including targets and backgrounds with stark contrast. On this basis, the chips online defect detection and classification algorithm including defects extraction and classification is researched. The defects extractions carried on the invariant and geometric positioning features of invariant moments to chips image correction and then use frame-difference method to extract defects. The defects classification is that a hybrid algorithm based on RBF Neural Network algorithm is applied to achieve the online detection for chips by real-time, fault-tolerant characteristics of neural networks. Above all, it satisfies the requirements of online detection.
PDF References Citation
How to cite this article
Zhang Daode, Xue Yangliu and Hu Xinyu, 2013. Study on the Flaw Recognition Algorithm Based on Neural Network. Information Technology Journal, 12: 7270-7274.
DOI: 10.3923/itj.2013.7270.7274
URL: https://scialert.net/abstract/?doi=itj.2013.7270.7274
DOI: 10.3923/itj.2013.7270.7274
URL: https://scialert.net/abstract/?doi=itj.2013.7270.7274
REFERENCES
- Poggio, T. and F. Girosi, 1990. Networks for approximation and learning. Proc. IEEE., 78: 1481-1495.
CrossRefDirect Link