Digitization of X-Ray Films of Aerospace Products and Defect Detection Based on Convolutional Neural Network

Abstract

The X-ray film digitization of aerospace products and the automatic detection of weld defects are of great importance for information storage and management. In this paper, we design a system for digitization and automatic detection. By automatically selecting the image with best exposure time among the images with different exposure time which are captured from an X-ray film, we can get the best digital image after homomorphic filtering of it. Then we design a network on the basis of YOLOv3 for defect detection. The digitization and detection results show that our system is of good effect.