Open AccessPerceptron Neural Network Based Machine Learning Approaches for Leather Defect Detection and Classification
Author(s) -
Praveen Kumar Moganam,
Denis Ashok Sathia Seelan
Publication year - 2020
Publication title -
instrumentation, mesure, métrologie/instrumentation mesure métrologie
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.161
H-Index - 4
eISSN - 2269-8485
pISSN - 1631-4670
DOI - 10.18280/i2m.190603
Subject(s) - artificial intelligence , artificial neural network , confusion matrix , pattern recognition (psychology) , multilayer perceptron , computer science , classifier (uml) , identification (biology) , confusion , automation , texture (cosmology) , perceptron , contextual image classification , machine learning , engineering , image (mathematics) , mechanical engineering , psychology , botany , psychoanalysis , biology
Detection of defects in a typical leather surface is a difficult task due to the complex, non-homogenous and random nature of texture pattern. This paper presents a texture analysis based leather defect identification approach using a neural network classification of defective and non-defective leather. In this work, Gray Level Co-occurrence Matrix (GLCM) is used for extracting different statistical texture features of defective and non-defective leather. Based on the labelled data set of texture features, perceptron neural network classifier is trained for defect identification. Five commonly occurring leather defects such as folding marks, grain off, growth marks, loose grain and pin holes were detected and the classification results of perceptron network are presented. Proposed method was tested for the image library of 1232 leather samples and the accuracy of classification for the defect detection using confusion matrix is found to be 94.2%. Proposed method can be implemented in the industrial environment for the automation of leather inspection process.