Real-Time Steel Inspection System Based on Support Vector Machine and Multiple Kernel Learning

With the higher quality standard from industries, the need for steel surface quality control has been greatly increased. The detection and recognition of steel surface defect is a critical issue for the quality control process. Among the techniques applied to tackle the problem, machine vision based approaches have advantages due to its flexibility, accuracy, and economy. This paper describes a real-time steel inspection system, which investigated the usage of support vector machine (SVM) and multiple kernel learning (MKL) method. Based on the preliminary experimental results, the proposed method demonstrates the efficiency in detection and recognition steel surface detects. It is shown that the advanced classification methods such as SVM and MKL are applicable for the real-time steel surface inspection system.