A Hybrid Feature-based SVM Approach for Image-based Fabric Type Classification

Accurately identifying fabric types is a complex and important issue in the textile industry, as materials such as cotton, denim, linen, polyester, and silk often have similar characteristics. Traditional manual methods of separating fabrics are time-consuming, expensive, and susceptible to human error. Recently, image-based fabric classification using computer vision and machine learning technologies has emerged as a promising solution. However, many conventional studies have focused only on texture- or colour-based features, which limits the ability to accurately distinguish between all types of fabrics. This study proposes a hybrid feature-based approach that integrates color features, shape features, and local binary pattern (LBP) texture features. By combining these hybrid features, we can represent the physical properties of each fabric in a multidimensional way, resulting in more accurate classification than achieved with a single feature type. The support vector machine (SVM) algorithm is used for classification and its performance is evaluated using different kernel functions, including radial basis function (RBF), linear and polynomial kernels. Experimental results show that the proposed method achieves very high performance, with an overall accuracy of 99.9%, precision of 0.9991, recall of 0.9991 and F1-score of 0.9991 in fabric classification, where the linear kernel exhibits superior performance in most cases. This study contributes to an automated image-based fabric recognition method that can be applied in quality control, inventory management and automated production processes within the textile industry.