Application Of Digital Image Processing In Quantitative Characterization Of Deformation Twins In TWIP Steel

Due to its high strength and good plasticity,twin-induced plasticity(TWIP)steel has a very potential application value in the automotive industry,ships,oil and gas development and other fields.The good mechanical properties of TWIP steel originate from the continuous high strain hardening caused by deformation twinning.Since the content of deformation twins directly affects the strain hardening ability of TWIP steels,the quantitative characterization of the integral fraction of deformation twins is of great significance for the study of TWIP steels.However,the traditional quantitative characterization methods of deformation twins mainly rely on manual measurement,the accuracy of measurement results is greatly affected by human factors,the measurement time is long,and it is not suitable for processing large-scale images.Aiming at this problem,based on the principle of digital image processing technology and the morphological characteristics of deformation twins,a digital image processing method was developed in this paper and applied to the quantitative characterization of the integral of deformation twins.It has the advantages of high accuracy and fast speed,which greatly improves the work efficiency.The microstructure images of the experimental TWIP steel under different strain states were observed and obtained using optical microscope and transmission electron microscope(TEM).On the basis of metallographic images and TEM images,combined with theoretical knowledge and tools such as image filtering,image enhancement,image segmentation,pattern recognition,etc.,the corresponding image processing methods and procedures were designed for the two microstructure images.Quantitative Characterization of Deformation Twin Fractions in TWIP Steels.For metallographic images,we developed an image processing method for block gradient segmentation.First use grayscale to convert color images to grayscale images,and then used bilateral filtering metallurgical smoothing filtering to remove noise while preserving edge information.Secondly,combined with the characteristics of deformation twins in metallographic images,a block gradient segmentation algorithm was designed to segment metallographic images.Then,the segmented binary image was processed with connected domain,and the noise of non-twinned features was removed according to the area size and shape of the connected domain.Finally,the area fraction of deformed twins in the metallographic image was determined by calculating the proportion of twinned pixels in the binary image.For TEM images,we developed an image processing method of Fourier direction filtering-Frangi direction filtering-threshold segmentation.According to the image characteristics,a contrast-limited adaptive histogram equalization method was first used for image enhancement.Secondly,an image segmentation method of Fourier direction filtering-Frangi direction filtering-threshold segmentation was designed to segment TEM images.The noise in the binary image was then removed using a manually selected method.Then,the deformation twins not identified in the previous process were extracted through image post-processing,and combined with the previous results,the area fraction of deformation twins in the TEM image was finally determined.Finally,combined with the manual measurement method-point counting method(TEM image),the metallographic image processing results and the advantages and disadvantages of the TEM image processing results were analyzed.Compared with the point counting method and TEM image processing results,the metallographic image processing results were larger,because the metallographic images represent twin beams,not single twins,but the optical microscope equipment was simple,and the metallographic images were easy to obtain.It wais used to analyze the approximate content and variation trend of twins.The results of TEM image processing are similar to those of the spot counting method,but the processing time of the TEM image was greatly reduced compared with that of the spot counting method,and the results were more stable and accurate.