Chemical testing and microscopic examinations are often used to identify wool. Wool identification in the textile business needs a method of efficient detection so that different varieties of wool may be properly categorized. Human eyes are used in the conventional textile Wool identification detection technique, which is both inefficient and costly. We are utilizing an ensemble of convolutional neural networks to increase the precision with which we can classify wool used in textiles. Our suggested study makes contributions to the field of wool prediction in the textile industry by proposing the usage of ensemble CNN for the identification of wool. These methods include quality prediction, feature extraction, and picture processing. Better feature extraction, Wool stem/core segmentation, feature matching, and identification are all possible because to image processing. RCB-Gray conversion, Noise reduction, and Contrast enhancement are the three main kinds of techniques used in image processing. Wool and hairiness features may be retrieved using feature extraction and classification; features were extracted using an ensemble CNN trained with a custom parameter set. Wool is classified into the high, low, and medium categories based on the quality forecast. Experimentation on the suggested models may now be evaluated with respect to error rate, processing time, prediction accuracy, and prediction, recall and F-measure.

Keywords: Wool parameters classification, CNN, SVM, feature extraction, and quality prediction
View Full Paper