The Study Of Axial Recognition Of Nitrogen-Vacancy Center In Diamond Based On Image Processing

Nowadays,quantum precision measurement technology has become one of the research hotspots in the application of quantum information.As one of the important branches of quantum precision measurement,detection of vector magnetic field has a wide range of applications in the fields of basic physics,biomedicine,and materials science.Nitrogen-Vacancy(NV)center is an excellent solid-state spin system,which has been extendedly applied in quantum precision measurement.Axial directions of NV center have different types,providing the theoretical support for its application in the detection of vector magnetic field.In addition,the accuracy of the axial information of NV center directly affects the precision of vector magnetic field measurement.In order to achieve the measurement of vector magnetic field,it is necessary to select three NV color centers with different axial directions in the diamond.Then,a threedimensional base vector is established in space to further obtain the vector information of the magnetic field.Therefore,to complete the calibration of the NV center axis under the laboratory coordinate system quickly and accurately,a fitting method based on the principle of angularly polarized beam imaging is proposed.Based on the image scanned from the angularly polarized beam,the identification and calculation of the axial information of the diamond NV center are realized by utilizing the theory of machine learning.Firstly,automatic identification and positioning of the diamond NV center are achieved based on a demonstration of the convolutional neural network model.Afterwards,on the basis of image processing,the extraction process of NV center axial information is optimized as well as the speed and accuracy of NV center fitting being improved,which contribute to reconstruction of vector magnetic field.Meanwhile,the proposed method in this design has some degree of robustness to the noise in the experimental environment.The work in this thesis mainly includes the following three parts:(1)Return difference elimination of the fluorescence image of NV center;(2)Object detection and positioning of the fluorescence scanning image of NV center;(3)Fitting and calculation of the azimuthal and polar angle of the axial information in the NV center target.In the traditional fitting method of NV axis,however,the accuracy of fitting algorithm about NV center fluorescence scanning image is low.Meanwhile,the algorithm is usually difficult to converge and need long time to fit.To solve this problem,an improved method based on accelerated image matching algorithm is developed.Firstly,the super-resolution generative adversarial network(SRGAN)is used to reconstruct the fluorescence scanning image of the NV center.After preprocessing the image,a suitable initial value of the fitting algorithm is obtained by combining the minimum enclosing area method with the image matching method.Eventually,the problem of non-convergence and falseconvergence is avoided and the accuracy of fitting is improved.Meanwhile,the fitting time is reduced and hence the efficiency of the experiment is finally improved.