Research On Quality Evaluation And Improvement Method Of X-ray Weld Image

This paper analyzes the characteristics of X-ray weld image,focusing on the characteristics of girth weld and spiral submerged arc weld image.In view of the poor background definition,blurred edge and even substandard blackness of the two X-ray weld images,the corresponding image quality evaluation,enhancement and noise reduction methods are studied.This paper takes girth weld image and spiral submerged arc weld image as the research objects,and discusses in the following aspects.(1)In the aspect of image quality evaluation,through the comparative analysis of subjective evaluation method,full-reference method and semi-reference method,human visual factors are considered in the quality evaluation.By analyzing the factors that affect human visual,such as illumination,area of region of interest(ROI),different weights are set for ROI and non-ROI,and an image contrast evaluation method considering visual factors is designed;By analyzing the relationship between image gray and image blackness,as well as the factors affecting image blackness,such as incident light intensity,projected light intensity and light source radiance,an image blackness quality evaluation index is proposed.The experimental results show that the above two evaluation methods are more accurate and more in line with the actual needs than the traditional full-reference and semi-reference methods.(2)In the aspect of image noise reduction,several targeted image noise reduction methods are proposed according to the image evaluation results.For the Gaussian noise filtering,several improved mean filtering algorithms are proposed based on the existing algorithms;For the salt and pepper noise filtering,median filtering is improved,and the noise recognition and filter window design are added.Compared with the existing algorithms,the improved algorithm has better robustness.(3)In the aspect of image enhancement,aiming at the problems of low contrast and uneven histogram distribution of weld image,several improved histogram equalization algorithms are proposed,and the enhancement effects of logarithmic function,exponential function and sin function are compared.Finally,a new weld image enhancement algorithm is proposed.It is proved that the proposed algorithm can automatically enhance the weld area instead of the background area,and does not need manual intervention in parameter selection.In addition,it has strong robustness,and can effectively enhance the girth weld X-ray image to the best personal visual perception value.(4)On the basis of the image quality evaluation and quality improvement methods,an automatic detection system for weld defects is developed,which includes X-ray weld image quality evaluation and enhancement.Compared with the existing detection methods,the success rate of weld defect detection is effectively improved.