Research On Key Algorithms Of Quantitative Metallography Based On Digital Image Processing

Quantitative metallography based on digital image processing is a subject that combines computer image processing with material science.Quantitative calculation of the parameters of metallographic microstructure by image processing method can quantitatively determine the mechanical properties of metal materials,which is of great significance to the quality detection and safety production of metal materials.This paper focuses on image filtering,edge detection,boundary restoration and reconstruction in metallographic image analysis.(1)Metallographic images are rich in texture and large in noise amplitude.Traditional image filtering methods are difficult to achieve ideal denoising effect.In order to solve this problem,the wavelet filtering method is introduced in metallographic image denoising,and the threshold function and adaptive threshold method of wavelet filtering are improved.Threshold function improves the malpractice that traditional wavelet filtering results may oscillate near the threshold or have a fixed deviation from the source image.Adaptive threshold guarantees that wavelet filtering can preserve image edge signal and remove noise in complex metallographic images.(2)Aiming at the problems of noise-sensitive,inaccurate edge location and bilateral edges,an edge extraction algorithm for metallographic images is proposed based on the improvement of image threshold segmentation algorithm.Experiments show that the algorithm not only has the advantages of precise location and good noise resistance,but also can mark inclusions in metallographic images while extracting edges.(3)Aiming at the common problems of edge loss and fracture of metallographic image and adhesion of different grains,the segmentation conditions of adhered grains are improved,and an image segmentation algorithm based on watershed algorithm is proposed.Experiments show that the algorithm achieves good results in metallographic image edge reconstruction and restoration,reduces the phenomenon of "over segmentation" and "wrong segmentation" in existing image segmentation algorithms,and makes the reconstructed edges clear,accurate and complete.(4)Aiming at the problems caused by traditional Zhang thinning algorithm,for example,the thinning result is not single-pixel width,and the source image topology will be changed,the principle is analyzed.Then,the labeling conditions of the classical Zhang thinning algorithm are modfied.The refinement result of the improved Zhang thinning algorithm is a single-pixel thin line with complete topology and smooth curve.(5)Aiming at the error of traditional chamfer distance transformation,the principle is analyzed,and it is proved that it is impossible to eliminate the error by changing the parameters.Then,the chamfer distance transformation algorithm is improved by replacing the distance information with the position information to realize the accurate Euclidean distance transformation.(6)Finally,the important parameters in traditional quantitative metallography and their calculation algorithm are analyzed,and the measuring methods of these parameters are redesigned based on the principle of digital image processing.Experiments show that the above algorithm has strong practicability in quantitative metallography.