Intelligent Detection of the Nanomaterials Spatial Structure with Synthetic Electron Microscopy Images

Abstract

One of the priority tasks in the process of chemical synthesis is study of nanoparticles morphological features. The key characteristics of materials at the molecular level are the size and geometric shape of the structure of the nanomaterials studied. One of the most effective methods for characterizing the morphology of nanoparticles is the transmission electron microscopy method. The main problem of applying this approach under the conditions of streaming synthesis of nanomaterials is the laboriousness and routine processing of electron microscopy images by a researcher. The creation of methods for automating the determination of the shape and size of nanoparticles will reduce the level of time and resource costs for diagnostics based on the results of the chemical synthesis of nanomaterials. This fact will positively affect the efficiency of scientific research in the field of synthesis and diagnostics of new functional nanomaterials. This paper proposes an accelerated method for generating synthetic images of transmission electron microscopy as a basis for creating software tools to accelerate the process of diagnosing and characterizing metal-organic frameworks in the process of chemical synthesis based on the use of deep learning technologies. The approach of automatic generation of the spatial structure of synthesized materials at the nanolevel is considered, which makes it possible to simulate a wide range of possible outcomes during laboratory synthesis. Procedures are proposed for generating sets of synthetic transmission electron microscopy images with support for automatic segmentation and extraction of areas of interest for machine and deep learning applications. The data sets obtained were tested and evaluated to solve the designated problem using the Yolo v5 object detection deep learning algorithm.