基于点光源X射线焊缝缺陷深度尺寸的测量

Measurement of Deep Sclae on Weld Defects based on Point X-Pay Source

关键词:  计算机视觉  无损检测  小波边缘检测

Key words:  computer vision  non-destructive testing  wavelet edge detection

摘要:

介绍了一种焊缝X射线图像缺陷深度信息实时自动检测方法。由于要实时和自动,则只能用经影像增强器转换的实时X光成像系统,而不能用照片方式。又由于现场条件的限制,不可能用多幅图的投影合成法。本文介绍了一种仅用三幅图来获取深度信息的方法。区别于照片成像,光源已经不能看成平行光,文中分析了点光源成像的原理和方法。二维缺陷的准确检测是计算深度信息的基础,而准确的边缘检测又是检测二维缺陷的基础。由Marr开创,经Canny、Witken和Mallat等人发展起来的多尺度边缘检测被认为是理论上最完善的一种方法,但他们没有解决尺度自适应问题。本文介绍了一种基于测度的尺度自适应边缘检测方法,获得很好效果。

Abstract:

This paper presents a real-time automatic method of deep information detection. A real-time image system with X-ray image-intensifier is used for requirement of real-time and automation instead of X-ray film. Number of image is limited due to lass image space. Only three projection images are used for 3-D information detection in this method. The light resource can not be booked upon a parallel light resource that is different form X-ray filme image approach. An image principle and approach based on point light resource is introduced. The detection of 2-D defects is based on computation deep information and precise edge detction is based on 2-D defects detection. A multi scales edge detection method, which was created by Marr and developed by Canny, Wiken and Mallat, is a perfect method. However, they could not solve the scale adaptive problems.A multi-scale adaptive method based on image information measurement is introduce in this paper.