Automated Scanning Electron Microscopy Image Analysis of Molybdenum–Iron Oxide Nanocomposites for Structure–Property Characterization Using Cellpose Segmentation and Machine Learning Classification

Structural properties of metal oxide nanocomposites play a crucial role in their performance in catalytic, environmental, and energy-related applications. This study presents an automated framework for the quantitative analysis of morphological and texture features of molybdenum-iron oxide nanocomposites using scanning electron microscopy (SEM) images. Instance segmentation was performed using the Cellpose deep learning model based on a U-Net architecture, enabling accurate separation of individual nanocomposites from dense and complex microstructures. Post-segmentation, 10 quantitative features, comprising six morphological descriptors (equivalent diameter, perimeter, aspect ratio, circularity, solidity, and eccentricity) and four Gray-Level Co-occurrence Matrix texture descriptors (contrast, homogeneity, energy, and correlation), were extracted. The resulting feature dataset was used to train four supervised machine learning classifiers: Random Forest, Support Vector Machine, eXtreme Gradient Boosting, and Light Gradient-Boosting Machine. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. The gradient boosting models demonstrated superior classification performance, with Light Gradient-Boosting Machine achieving the highest mean accuracy of 95.1% for shape classification, while eXtreme Gradient Boosting achieved the highest mean accuracies of 96.9% for size range classification and 94.8% for texture classification. These results highlight the effectiveness of gradient boosting techniques in capturing complex relationships between morphological and texture features for automated nanocomposite characterization. Overall, the proposed framework enables efficient automated molybdenum-iron oxide SEM image analysis and quantitative nanocomposite characterization, providing deeper insight into structure-property relationships in advanced nanomaterials.