基于机器视觉的量子点STM形貌图像识别研究

为减轻量子点表面形貌分析过程中的人工工作,使量子点的STM图像分析更加自动化,基于机器视觉对衬底的斜切角及量子点的形貌特性展开研究.利用腐蚀和边缘检测提取台阶形状,并通过反三角变换计算斜切角.利用二值化和阈值下降对量子点的数量与空间坐标进行提取,在此基础上,通过邻域密度计算分析其均匀性,并在解决图像中的粘连问题后找出量子点的尺寸.实验结果显示,与人工统计相比,斜切角、量子点计数及尺寸的平均误差分别为5.02%, 0.7788%及1.12%;并实现量子点均匀性的自动化统计与分析.基于机器视觉算法的自动识别过程,对协助研究者分析量子点表面形貌有实际意义.

Research on Quantum Dot STM Morphological Image Recognition Based on Machine Vision

In order to alleviate artificial work in the process of surface topography analysis of quantum dots and make the STM image analysis process more automation, substrate miscut angle and morphological characteristics of quantum dots are investigated based on machine vision. Firstly, the step shape is extracted by erosion and edge detection, thus the miscut angle is calculated by inverse triangulation. And then, quantum dot number and corresponding spatial coordinates are extracted by binarization and threshold descent, based on which its uniformity is calculated by the neighborhood density, and size is found out after solving adhesive problem in image. Experimental results show that the average error for miscut angle calculation, quantum dot number and size statistics is 5.02%, 0.7788% and 1.12% respectively compared with the manual method; and further provided the automotive capability for quantum dot uniformity statistics. The automatic recognition process based on machine vision algorithm has practical value for assisting researchers to analyze surface morphology of quantum dots.