Research On Key Technology Of Weld Defect Detection System Based On X-ray Image

In the field of industrial weld inspection,various defects in welding can cause many dangerous accidents.At present,in the non-destructive welding seam inspection,the main method for evaluating the welding quality is manual inspection.Manual detection has problems such as inconsistent subjective standards,low detection efficiency,and complex tasks.Therefore,the automatic identification of X-ray weld image defects has very important research significance.In the X-ray weld defect detection and recognition system,the weld image is pre-processed first.In view of the characteristics of noisy and uneven distribution of the X-ray weld image,this paper uses an improved algorithm of wavelet transform,the algorithm Overcame the shortcomings of the traditional wavelet transform,and the weld image after noise reduction was low in contrast,and the histogram of the X-ray weld image was analyzed,so an improved double histogram equalization algorithm was used to enhance the contrast Finally,for the X-ray weld image,there are a lot of backgrounds that are not related to the weld area,and the defect only exists in the weld area,so the improved three-dimensional Otsu algorithm is used to segment the weld area in the X-ray weld image.After experiments and verification,the algorithm used in this paper can effectively reduce the noise of the weld image,enhance the contrast of the weld image,and accurately segment the weld area from the entire image.Defect identification plays an important role in the automatic identification of X-ray weld defects,and defect identification is the basis for evaluating welding quality.Since there are five types of defects to be detected in this paper,and the traditional support vector machine can only perform two classifications,this paper uses an improved binary tree algorithm,which first combines the binary tree with the support vector machine,and then proposes A new strategy that solves the inheritance of traditional binary trees.The experimental results show that the improved binary tree algorithm has a higher defect recognition rate and runs faster.