Research On TEM Nanoparticle Parameter Measurement Method Based On MaskR-CNN Segmentation

The characteristics of nanomaterials are related to the size and morphology of nanomaterials.According to the nanoparticle image captured by transmission electron microscopy(TEM),one can not only measure the size of particles,but also observe the distribution of particle size and particle morphology.To improve the efficiency,accuracy and automation of the measurement for nanoparticle size,image processing technology can be used to process and analyze the TEM image of nanoparticles.In this thesis,Mask R-CNN network was used in edge segmentation for spherical and rod-shaped nanoparticle images.On this basis,edge fitting was carried out for the segmented nanoparticles.Finally the size of the nanoparticles was measured and the automatic analysis and evaluation for particle size were realized.The main work of this thesis includes the following aspects.Firstly,in view of the characteristics of agglomerated adhesive particles in the nanoparticle images,Mask R-CNN network was selected to segment the edge of nanoparticles,and the network was optimized to improve the accuracy of segmentation.Secondly,according to the particle segmentation results,the boundary of the nanoparticles is fitted.For the complete nanoparticles and incomplete nanoparticles,different fitting methods are adopted to fit the boundary of the nanoparticles into basic geometric shapes similar to their shapes.Finally,the size and shape parameters of particles were measured based on the fitting results,and the data were statistically analyzed.The uniformity of particles was evaluated according to the statistical results.The experimental results verify the effectiveness of the proposed nanoparticle size measurement method based on Mask R-CNN segmentation.