Download PDFOpen PDF in browser

Defect Detection for Wet-laid Fiberglass Mat Based on An Automatic Iterative Otsu Method

EasyChair Preprint 1076

8 pagesDate: June 1, 2019

Abstract

Real-time defect detection in industrial products is a challenging problem. Image segmentation is an extremely important step in defect detection. Threshold is one of the most commonly used methods in image segmentation, which can segment objects from image background. As a classic thresholding method, Otsu can obtain satisfactory threshold value when the difference between object and image background is obvious. However, when the proportion of the object is small, Otsu method has a poor segmentation effect. In this paper, we propose a defect detection method based on Otsu method to tackle the problem of automatic defect detection of wet-laid fiberglass mat. After the pre-processed image is segmented by the Otsu method, we remove the isolated pixels in the image and mark the remaining image areas. The marked image area will be processed as a new image. Then we repeat the above operation until the defect location is accurately determined. The proposed automatic iterative Otsu method can continuously reduce the size of the image to be processed, increase the proportion of defect area and improve the image segmentation effect of Otsu method. The experimental results show that our method can largely eliminate the image noise, determine the defect location and achieve high robustness and computational efficiency in automatic defect detection.

Keyphrases: Otsu method, defect detection, image segmentation, wet-laid fiberglass mat

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
@booklet{EasyChair:1076,
  author    = {Chengming Zou and Ronggang Xue and Xiaolong Xu},
  title     = {Defect Detection for Wet-laid Fiberglass Mat Based on An Automatic Iterative Otsu Method},
  howpublished = {EasyChair Preprint 1076},
  year      = {EasyChair, 2019}}
Download PDFOpen PDF in browser