Welding Seam Defect Detection And Recognition Method Based On X-ray Image

In the welding process,affected by many factors,it is easy to form weld defects,which affect the strength and corrosion resistance of the weldment.In order to ensure product quality and prevent accidents,the detection of weld defects is very important.At present,the method of manual evaluation of weld defect images is mainly used.This method is limited by the subjectivity of the inspectors,and there will be problems of missed detection and false detection.The deep learning model can automatically extract image features and realize defect location and classification,which provides a new direction for weld defect detection.After analyzing the structure of the existing deep learning model,a welding defect detection and identification method based on the improved SSD model is proposed.Aiming at the problem of noise interference during the acquisition and transmission of X-ray weld images,median filtering is selected as the image noise reduction method.In order to enhance the contrast between the weld area and the background area and highlight the key feature information,the histogram equalization method is used to realize image enhancement processing.The algorithm principle of the main target detection model is analyzed,and the VOC2007 public dataset is used to conduct model performance comparison experiments,and finally the SSD model is selected as the basic framework.Combined with the actual needs of weld defect detection and identification tasks,a series of improvements are made to the SSD model.Use Res Net-50 as the model backbone network to improve the feature extraction ability,and use the residual block structure to solve the problem of training gradient disappearance;add a feature fusion module to make up for the lack of insufficient shallow feature information,and improve the model’s small target detection ability;add an attention module to improve the model’s attention to important information.A dataset of X-ray weld images containing five typical defect types is constructed.Under the same experimental conditions,the performance of the SSD model and the improved models at different stages are tested.Compared with the original SSD model,the detection accuracy of the improved RAF-SSD model is significantly improved,which has practical value in the field of weld defect detection and identification.Figure 45;Table 6;Reference 53...