EBSD花样菊池带的快速识别

Fast Identification of Kikuchi Bands in EBSD Patterns

电子背散射衍射(electron backscatter diffraction,EBSD)领域引入Hough变换,实现了由计算机自动识别EBSD花样中的菊池带,EBSD技术由此发展成为表征晶体材料取向的一种重要手段。由于Hough变换的速度远远慢于扫描电子显微镜中电子束的移动速度,如何快速实现Hough变换已成为整个EBSD领域亟待解决的关键问题。分别利用中央处理器(central processing unit,CPU)和统一计算设备架构(compute unified device architecture,CUDA)对EBSD花样做Hough变换,详细比较了两者的耗时和识别结果,测试表明:CPU对EBSD花样菊池带的最快识别速度为1.7×10^(3)FPS,同样条件下,CUDA对EBSD花样菊池带的识别速度最快可达1.7×10^(4)FPS,远远超过CPU的最快识别速度,且利用CUDA进行Hough变换并没有影响EBSD花样菊池带的识别效果。基于CUDA的Hough变换将成为快速识别EBSD花样菊池带的一个发展趋势。

In the field of electron backscatter diffraction(EBSD), Hough transform was introduced to realize the automatic identification of Kikuchi bands in EBSD patterns by means of computer, and EBSD technique has been developed into an important means to characterize orientation of crystalline materials. Since the speed of Hough transform is much slower than the moving speed of electron beam in scanning electron microscope, how to carry out Hough transform quickly has become a key problem that needs to be solved urgently in the whole EBSD field. In the present work, central processing unit(CPU) and compute unified device architecture(CUDA)were used to perform Hough transform on EBSD patterns, respectively, and the time consumption and recognition results were compared. The results show that the fastest recognition speed of Kikuchi bands is 1.7 ×10^(3)EBSD patterns per second for CPU. Under the same conditions, CUDA can process 1.7×10^(4)EBSD patterns per second,which is far faster than the fastest recognition speed of CPU. More importantly, Hough transform based on CUDA does not affect the identify results of Kikuchi bands in EBSD patterns. As a result, Hough transform based on CUDA technique will become a trend for fast identification of Kikuchi bands from EBSD patterns.