Research on Intelligent Fabric Texture Classification Based on Mathematical Modeling and Image Processing

Fabric texture recognition is a critical component in the automation of the textile industry, placing high demands on the perception and classification of complex texture features. To address the limitations of traditional methods—namely, low recognition accuracy and poor robustness in multi-category scenarios—this study develops an intelligent fabric texture classification system that integrates image processing with mathematical modeling. Experimental results demonstrate that the improved Principal Component Analysis-Fusion Convolutional Neural Network-Support Vector Machine (PFCS) algorithm achieves an average recognition accuracy of 94.2%. Notably, the classification accuracies for denim and coral fleece reach 96.4% and 95.9%, respectively. Five-fold cross-validation reveals a minimum model accuracy of no less than 91.7% across different data partitions, indicating strong generalization capability. This research offers a viable approach for high-precision, intelligent classification of multi-category fabric textures and exhibits significant potential for engineering applications.