Research On The Surface Morphology Of InGaAs Quantum Dots Based On Machine Vision

Self-organized quantum dots have attracted the attention of many researchers because of their special electrical and optical properties.However,the process of surface topography analysis of quantum dots requires human participation.In order to make the analysis process more automated,the method of machine vision is applied to the process.First,the image stitching technology in machine vision was used to study the stitching method of quantum topography images.When using scanning tunneling microscopy or atomic force microscopy to characterize the surface structure of quantum dots,the problem of obtaining a scene with a large viewing angle while retaining fine textures was studied.Then,the parameters of the substrate's miscut angle and the number,density,size and uniformity of the quantum dots,which have important effects on the performance of the quantum dot device,were calculated by the method of machine vision,making the analysis of the surface topography of the quantum dots more automated.The main contents are as follows:(1)For the stitching of quantum images,there is two problem that the Harris algorithm needs to manually set the threshold and the quantum image similarity is large.A quantum image stitching algorithm based on improved Harris and quadratic normalized cross correlation(NCC)is proposed.In terms of threshold setting,due to the large local similarity of the image,the image is first divided into 3 � 3 subgraphs;then the number of quantum dots and rings of a subgraph is calculated based on binarization and threshold descent.Finally,the threshold of the sub-picture is determined by the number,which is also the overall threshold,so as to obtain an appropriate number of corners.In terms of large image similarity,first select a smaller window for NCC corner matching,and then select a larger window to perform the second NCC calculation on the basis of the previous one to reduce the mismatch rate.The experimental results show that the improved algorithm has better effects on speed and accuracy in the calculation of quantum dots and rings;in terms of threshold setting,the number of corners detected by the improved algorithm is more reasonable than the traditional algorithm;in terms of large image similarity,the improved algorithm has a faster matching speed and a lower mismatch rate,and the mismatch rate are 4.82% ? 27.27%.Therefore,this algorithm effectively improves the reliability and speed of quantum image stitching.(2)To solve the problem of unreasonable number of feature points and high mismatch rate in the traditional scale-invariant feature transformation(SIFT)algorithm when stitching quantum structure topography images,a SIFT quantum image stitching algorithm based on dynamic threshold and global information is proposed.In terms of the number of feature points,based on the characteristic that the number of quantum dots and rings have a direct impact on the number of feature points,the density of quantum dots and rings is used to dynamically set the contrast threshold to ensure that the number of feature points is within a reasonable range.In terms of the mismatch rate,the concentric circle analysis domain is constructed to generate descriptors of global information,and combined with the local information descriptors detected by the SIFT algorithm to reduce the mismatch rate of the algorithm.The experimental results show that the number of feature points detected by the improved algorithm is more reasonable than the traditional algorithm.And the improved algorithm reduces the mismatch rate to 7.59% ? 28.83%.Therefore,the improved algorithm has better reliability in quantum image stitching.(3)In order to reduce the artificial participation in the analysis of the surface morphology of the quantum dots and the process of analyzing the image of the topography of quantum dots is more automated,the miscut angle of the substrate and the morphological characteristics of the quantum dots were studied based on machine vision.First,the step shape is extracted using methods such as erosion and edge detection,and the miscut angle of the substrate is calculated using inverse triangulation.Then,the number and density of quantum dots are counted through binarization and threshold reduction.On this basis,the uniformity of quantum dots is analyzed through neighborhood density calculation.After solving the adhesion problem in the image,statistics the size of quantum dots.The three-dimensional modeling method is used to calculate the crystal plane index of the quantum dots.Experimental results show that,compared with artificial statistics,the average calculation error of the machine vision method in terms of substrate chamfer angle,quantum dot number,size and crystal plane index are 5.02%,0.7788%,1.12% and � 0.249,respectively analysis.Automated statistics and analysis of the uniformity of quantum dots are achieved.Therefore,the automatic recognition method based on machine vision has potential application value to assist researchers in analyzing the surface morphology of quantum dots.