Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


16152.pdf (9.03 MB)

ETD Abstract Container

Abstract Header

Automated Defect Recognition in Digital Radiography

Xiao, Xinhua
http://orcid.org/0000-0003-3401-2176
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439309683

Abstract Details

2015, PhD, University of Cincinnati, Engineering and Applied Science: Electrical Engineering.
Digital radiography has been widely used for non-destructive testing in industrial production. An example is the inspection of industrial parts like turbine blades of jet engines. ADR (Assisted Defect Recognition) is an X-ray image based inspection system developed for anomaly detection of turbine blades. The system uses a reference-based method, in which statistical models are created at pixel level based on a reference model image set from good parts. And pixels of the images under test, when compared with the statistical models, which yield significant differences are called out and identified as potential defects. The system is efficient to detect low-contrast defects, but its effectiveness heavily relies on the reference model image set. When non-representative reference image sets are used, there is a high probability of false rejections. Due to variations in the production process, the reference image set may have to be adapted. The research work proposes an automatic approach to select, based on the feature extractions of callout images of the system, a representative reference model image set for the system. Experimental results show that the proposed approach can select a model image set with a low false alarm rate and acceptable detection rate and outperforms manual approach. To adapt to the variations in the production process, an adaptive procedure based on the automatic approach is proposed to update the reference model set. Experimental results show that the proposed procedure can automatically detect significant variations and update the model set with little human intervention. The research work also studies the impact of using reference model image sets containing images of parts with defect indications (imperfect images). A systematic procedure is proposed to evaluate the impact of imperfect images on the performance of ADR based on McNemar's test. The number of imperfect images which can be tolerated in the model set is determined for each type of defect indications. To further improve the defect recognition rate, the research work proposes a hybrid method by combining the ADR system and a new classifier based on scan line and modified Haar-like features. An image under test is first inspected by the ADR system. If the image is not called out by ADR, scan line and modified Haar-like features are extracted at a series of specific regions. If the decision rules defining the normal feature space which are learned from the model image set are not satisfied, the image is considered as defective. Experimental results show that the proposed method can detect all non-callout strong defective images by ADR without increasing false alarm rate. The proposed method is also applicable to the detection of strong positive images.
William Wee, Ph.D. (Committee Chair)
Raj Bhatnagar, Ph.D. (Committee Member)
Chia Han, Ph.D. (Committee Member)
Anca Ralescu, Ph.D. (Committee Member)
Xuefu Zhou, Ph.D. (Committee Member)
103 p.
Electrical Engineering
reference image selection; statistical test; image classification; turbine blade; model adaptation; X-ray

Recommended Citations

Citations

  • Xiao, X. (2015). Automated Defect Recognition in Digital Radiography [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439309683

    APA Style (7th edition)

  • Xiao, Xinhua. Automated Defect Recognition in Digital Radiography. 2015. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439309683.

    MLA Style (8th edition)

  • Xiao, Xinhua. "Automated Defect Recognition in Digital Radiography." Doctoral dissertation, University of Cincinnati, 2015. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439309683

    Chicago Manual of Style (17th edition)

ucin1439309683
1,306
© 2015, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.