Date of Publication :15th June 2024

Abstract:The precise classification of fabrics as either synthetic or natural is crucial for various industrial and consumer applications, ranging from textile recycling to healthcare product development. This study investigates the efficacy of an ensemble of image processing filters combined with machine learning models in achieving accurate fabric classification. We propose a novel approach that leverages the complementary strengths of various image processing techniques - namely, Local Binary Pattern and Gabor filter - to extract textural features from fabric images. These features are then fed into distinct machine learning models - namely, SVM, Naive Bayes Classifier, Random Forest, CatBoost, and XGBoost - to perform the classification task. The performance of each model is evaluated individually without the filters and with filters, using a comprehensive dataset of fabric images encompassing a wide range of synthetic and natural materials. Our findings demonstrate that the proposed combination of image filters significantly impact the performance of specific classification models. This research paves the way for the development of more robust and reliable fabric classification systems with potential applications in diverse fields.

Reference :