Research On Counterfeiting Detection Of Weld Seam Radiographic Inspection Image

Welding technology is an important processing technology in manufacturing industry,which is widely used in aerospace,chemical production,pipeline transportation and other aspects.In the welding process,due to the negligence of the staff or the influence of the working environment,there will be some defects in weld,such as cracks,gas holes,missing area and so on.In practice,tampering with the weld image often happen,for example,by replacing a defective weld image directly with a qualified weld image,or by re-taking a qualified weld and replacing the label information.When the reader detects the defect of the weld image,he will also spend a lot of energy to detect whether the overlap of the two weld images can be successfully overlapped to judge whether the image has been replaced or tampered This method is time-consuming and inefficient,which is not conducive to staff to improve production efficiency.So three methods of automatic detection are proposed in this paper to help workers detect effectively.(1)A tamper detection method based on overlap matching of weld seam.The tamper operation of directly replacing weld image will lead to the overlap area of adjacent two weld images can not match successfully,so this method use SIFT algorithm and SSIM algorithm to find the matched key points of overlap mark and overlap seam between two weld image and detect the structural similarity of the overlap seam.According to the detection results of the two algorithms,we can judge whether the overlap area can be matched successfully,then judge whether the image is tampered.(2)A tamper detection method based on tag information.The lable information in weld image is to record the time of shooting,welder number,pipeline number,center mark,lap mark and so on.The lable information only identifies a weld image.Therefore,this method use the convolutional neural network to identify the tag information of the image to be detected.By comparing the tag information with the stored tag information in the database,the duplication of the image can be judged.(3)A tamper detection method based on weld texture features.Different weld seams have different texture features.If the someone re-shoots the qualified weld and replaces the label information,it will lead to repeated texture features in the database.The method extracts the seam texture features from the image to be detected to find the seam image with similar texture features,and detects whether the seam exists bycombining SIFT algorithm.