Image Mosaic And Defect Detection Technology Of Small Pipe Surface Defects After Fluorescent Penetration

In order to effectively detect the typical surface defects of small pipes and improve the detection efficiency of small pipes manufacturing surface defects,this article was titled "Image Mosaic and Defect Detection Technology of Small Pipe Surface Defects after Fluorescent Penetration",focusing on the research of image registration,image fusion and semantic segmentation technology based on fluorescence penetration method,developing a small pipe manufacturing surface defect detection test device to realize those functions,which had important practical significance and academic value to promote the automation,unmanned and intelligent level of surface defect detection in pipe fitting manufacturing.The research was supported by science and technology project of Guangdong Institute of Special Equipment Inspection and Research in 2019(2020CY14).This paper analyzed the research progress of related content at home and abroad from the aspects of fluorescent penetration image mosaic method and penetration detection method of small pipe manufacturing surface defects,and then determined the research content of this paper.The main contents of this paper outlined as follows:(1)The technical framework of fluorescence penetration image mosaic and detection of surface defects in small pipe manufacturing was designed.Based on the analysis of the overall requirements of fluorescent automatic penetration detection for surface defects of small pipes manufacturing,the mechanism of fluorescent penetration detection for surface defects of small pipes manufacturing was studied.It was pointed out that the realization of fluorescent penetration image mosaic and detection for surface defects of small pipes must solve the problems of fluorescent penetration image based on homography transformation and fluorescent penetration image mosaic registration technology based on feature extraction The key technologies such as splicing fusion technology,the realization of the surface defect evaluation device based on deep learning of small pipes was discussed.(2)An improved method of homography transform fluorescence penetration image mosaic registration was proposed,based on the lateral surface unfolding.On the basis of the limitations of the classical homography transform image mosaic registration method,the registration process of small pipe fluorescent penetration image under different angles was analyzed,and the influence of the surface expansion model of small pipe on the process was studied.The improved homography transform based fluorescent penetration image mosaic registration method was used to carry out the mosaic registration experiment of small pipe fluorescent penetration image under different angles.(3)An image mosaic fusion method of small pipe fluorescent penetration based on feature extraction was proposed.This paper introduced the method of fluorescence penetration image fusion based on feature extraction,analyzed the distribution law of the brightness features of each angle of fluorescence penetration image,adjusted the outlier component of brightness index,improved the fusion effect of fluorescence penetration image,and comprehensively evaluated the quality of fluorescence penetration image stitching and fusion.(4)Surface defect evaluation and experiment of small pipe manufacturing based on deep learning was designed.Based on the deep learning convolution network,the image features were extracted,the loss function was calculated,the pixel coordinates and mask size were determined,and the quality level was evaluated.The experimental device for stitching and detecting the surface defects of small pipe fittings was built,and the software for stitching and detecting the surface defects of small pipe fittings was made.The results showed that the surface defect detection technology and test device based on deep learning were effective and feasible.