Feature extraction with an associative neural network and its application in industrial quality control

Abstract

There are several approaches to quality control in industrial processes. This work is center in artificial vision applications for defect detection and its classification and control. In particular, we are center in textile fabric and the use of texture analysis for discrimination and classification. Most previous methods have limitations in accurate discrimination or complexity in time calculation; so we apply parallel and signal processing techniques. Our algorithm is divided in two phases: a first phase is the extraction of texture features and later we classify it. Texture features should have the followings properties: be invariant under the transformations of translation, rotation, and scaling; a good discriminating power; and take the non-stationary nature of texture account. In Our approach we use Orthogonal Associative Neural Networks to Texture identification and extraction of features with the previous properties. It is used in the feature extraction and classification phase (where its energy function is minimized) too, so all the method was applying to defect detection in textile fabric. Several experiments has been done comparing the proposed method with other paradigms. In response time and quality of response our proposal gets the best parameters.