Fast Automatic Accurate Identification Of The Kikuchi Bands In Electron Backscattered Diffraction Pattern

Electron Backscatter Diffraction(EBSD)has become a standard accessory for scanning electron microscopy(SEM).With the Kikuchi pattern obtained by EBSD,SEM can perform phase identification,crystal orientation and texture analysis.The existing applications are technically mature,and the algorithm is mainly based on Hough transform to identify Kikuchi bands of EBSD pattern.Since Hough transform could not recognize hyperbola directly,the information obtained from the traditional Hough transform was inaccurate,which had a fatal effect on using EBSD pattern to analysis Bravais lattice of unknown crystal.For this reason,this paper mainly discusses the algorithm of accurately identifying the hyperbolic boundary of Kikuchi bands.At the same time,it uses multi-threading technology to realize the whole identification process quickly and automatically,and develops an application program.Firstly,the Hough transform was performed on the EBSD pattern,and the Hough transform results were used to identify the Kikuchi bands with the most obvious contrast in the EBSD pattern.Then,the traces of these Kikuchi bands were Fine-tuned and optimized.The gray distribution curve was statistically analyzed along the hyperbola,and the recognition accuracy was improved according to the principle that the boundary of Kikuchi band had the maximum gray gradient.Based on the inherent constraint law between crystal surface and crystal direction in crystallography,the traces are repeatedly fitted by the intersection points which on the same traces of Kikuchi bands,and the accuracy of the intersection points is further improved based on the optimized traces.More Kikuchi bands are identified by these connecting line of intersections,and binding the Kikuchi bands to improve the overall identification accuracy of Kikuchi bands.Based on the above algorithm,the application software is developed by using the parallel computation of CPU and GPU to speed up the whole recognition process.