An Automatic System of Surface Defect Classification of Hot Rolled Steel

(1)

The 3^rdInternational Conference on Electromechanical Engineering ICEE’2018, November 21-22, 2018,Skikda Mechanical Engineering and materials

1

An Automatic System of Surface Defect Classification of Hot Rolled Steel

Adel Boudiaf

^{1, a*}

, Oussama Kadri

^{2, b}

,Rachid Zaghdoudi

^1,c

, Zoheir Mentouri

^{1, d}

,Samira Taleb

^{1, e}

1Research Center in Industrial Technologies CRTI P.O.Box 64, cheraga 16014 Algiers, Algeria

2Electrical engineering department, Mohamed Khider University, 07000 Biskra,Algeria

ABSTRACT—L'objectif principal de cet article est de développer un système d'inspection automatique des défauts de surface des aciers plats laminés à chaud. La technique proposée comprend quatre étapes principales. La première étape consiste à l’acquisition d’image. La deuxième étape estl’extraction decaractéristiques de l'image par histogramme de gradients orientés (HOG). Dans la troisième étape, l'analyse en composantes principales (ACP) est appliquée au descripteur HOG afin de réduire la dimensionnalité du vecteur de caractéristiques. En fin, le classificateur KNN (K-proche voisin) est utilisé pour classifier les différents défauts de surface de l'acier. Les résultats expérimentaux ont montré que le système d'inspection d'acier basé sur le classifieur KNN proposé fournit de meilleurs résultats, avec une précision d'environ 91,12%

MOTS-CLÉS :Vision industrielle; le classificateur K- le plus proche voisin; classification des défauts de surface;

Histogramme des gradients orientés (HOG); Analyse en composantes principales (ACP).

ABSTRACT. The main objective of this paper is to develop an automatic surface defect inspection system for hot- rolled flat steel. The proposed technique consists of four major steps. The first step is image acquisition. The second step is featured an extraction of the image by Histogram of oriented gradients (HOG). In the third step, the principal component analysis (PCA) is applied for the HOG descriptor to reduce the dimensionality of the feature vector. In the final step, the K- nearest neighbor classifier (KNN) is used to classify the different steel surface defects. The experimental results showed that the proposed inspection system of steel based on KNN classifier provides better results, whilst achieving accuracy about 91.12%.

KEYWORDS:Industrial vision; the K- Nearest Neighbour classifier; classification of surface defects; Histogram of Oriented Gradients (HOG); Principal Component Analysis (PCA).

1. Introduction

Today, products of rolling represent approximately 90% of all materials production in the sector of metalworking because these rolling products are widely used in different fields of industry such as a food packaging, military, medical, auto industry, aerospace, and other fields [1].

Many of these industries (aerospace and automotive) reject the rolling products which have surface defects because a minor defect in these materials might result in Human and material losses in later stage [1]. Studies [2] show that 70%

of the failures of rotating machinery are due to the surface defects as shown in Figure.1. Therefore, these surface defects must be detected in early stage as possible as to avoid the loss of production, and also preserve the safety of personnel.