Defect detection in textiles using back propagation neural classifier

The textile products are affected by the defects during the manufacturing processes. It is also waste of the resources used for the production and in turn it affects the business. The manual inspection in defect detections is not encouraged these days in manufacturing process. The computer vision with machine learning algorithms in automated quality control system plays an important role in detecting defects in manufacturing process as well as analyzing the quality of products. Classification of defects in knitted fabric is an active area of research around the globe. This paper presents a classification method to detect defects such as holes and thick places in knitted fabric by applying artificial neural network algorithm. The artificial neural network algorithms learn from the input data after successful training process, it predicts the nature of the unknown samples in very fast and accurate way. The proposed work has been carried out in two phases. In the first phase the images of the defective samples of two classes were collected by a high-resolution camera. The color images of the samples were converted into grey scale images. The features were extracted from each grey scale image and stored in a database. In the second phase a neural classifier was trained with back-propagation neural Network (BPNN) algorithm on the training dataset. After successful training of the neural network on train dataset, the performance of the trained neural network was evaluated on the test dataset. Different experiments were carried out by increasing the number of training data samples; it was found that the best evaluation performance was obtained as 83.3%.