False film identification method of industrial radiographic film

Abstract

The invention relates to the technical field of deep learning and image processing, in particular to a false film identification method for an industrial radiographic film. The false film identification method comprises the following steps: inputting a reference film and a film to be identified, firstly, respectively extracting weld joint areas of two negative films by using a Gaussian difference method twice, extracting five areas of interest which comprise a central point and are symmetrical about the central point of the weld joint from the weld joint areas, and carrying out pseudo-color processing on the areas of interest; secondly, sending the areas of interest subjected to pseudo-color processing into an improved twinning neural network for training and recognition to obtain numerical values after pairwise comparison; and finally, establishing a similarity judgment system, carrying out self-defined weight operation, and judging that the film is a false film if the weight is greater than a set threshold value. Experiments prove that the method can effectively distinguish the false films in the industrial films, the accuracy rate can reach 99.1% or above in 261 test set samples, and the method is simple and effective and has certain practical feasibility.

Description

technical field [0001] The invention relates to the technical field of deep learning and image processing, in particular to a method for identifying fakes of industrial radiographic negatives. Background technique [0002] With the rapid development of welding technology, the quality of welding has become an important concern for the manufacturing industry to judge welding safety. When the number of welds is too large, some welds may not be photographed or one weld may be photographed multiple times during the digital imaging of welds, resulting in industrial fake images, resulting in a large number of welds not being evaluated. Cause huge security and economic losses. [0003] The fake recognition problem in industry is the similarity problem of welds in two negative images. At present, the industry generally adopts the working method of distinguishing "fake films" with human eyes, which is time-consuming, labor-intensive, heavy workload, cumbersome and low accuracy; [...

AI Technical Summary

Problems solved by technology

[0007] In view of this, the present invention provides a fake identification method for industrial radiographic negatives in order to solve the various deficiencies of existing artificial means and the similarity evaluation and low accuracy of other current methods for such negatives with large aspect ratios. , which ensures accuracy and efficiency by constructing a twin neural network, then optimizing the data set and improving the backbone feature extraction network

Images