The primary objective of this study is to classify defective Adire fabrics, allowing quality control to prevent economic loss. In Adire production, common defects include stains, tears, colour variation, colour smear, colour bleeding, shade variation, pattern misalignments and colour crocking. Defect classification in Adire production is critical for ensuring high product quality are churned out to meet the needs of end users and reduce wastages. The use of Machine Learning models to automatically classify defective Adire fabrics proffer solution to the ineffective and time-consuming manual methods. A total of 234 Adire fabric images were captured at Itoku Abeokuta, Ogun state with a Redmi 14C 50MP digital camera and augmented to 884. Among the data collected, 396 images represented defective while 488 were non-defective (Normal). Preprocessing involved Gaussian filtering and Contrast Limited Adaptive Histogram Equalisation (CLAHE), while Gray-Level Co-occurrence Matrix (GLCM) was used for texture-based feature extraction. SBOA was applied to optimise SVM hyperparameters (penalty factor C, kernel type, gamma (Y), and polynomial degree) yielding the SBOA-SVM model, implemented in MATLAB R2023a. The visible results obtained from the optimised model in comparison with standard SVM confirms that SBOA-SVM model outperformed the standard SVM across all evaluation metrics. The aggregate classification accuracy improved from 92.87% (SVM) to 94.68% (SBOA-SVM), Sensitivity increased from 91.67% to 93.69%, Specificity also rose from 93.85% to 95.49%, while the false positive rate decreased from 6.15% to 4.51%, confirming fewer normal fabrics were incorrectly classified as defective. Importantly, the computation time of SBOA-SVM model (116.22 s) was slightly lower than that of the standard SVM (122.36 s) without additional computational burden. These findings underscore SBOA-SVM potential for replacing manual inspection processes in real-world fabric inspection processes, where precision and speed are critical.
Ogunleye et al. (Thu,) studied this question.