LAUSR

Feature extraction and classification recognition are very important for surface defect recognition of strip steel. Traditional methods of feature extraction and defect recognition cannot take into account both the recognition accuracy and the running time. In order to solve the above two problems, a novel feature extraction and recognition method are proposed in this paper. On the one hand, a semi-supervised principal component analysis and locality preserving projection manifold learning algorithm are proposed to reduce features dimension. On the other hand, the particle swarm optimization–second-order cone programming–multi-kernel relevance vector machine (PSO–SOCP–MKRVM) classification algorithm is proposed in this paper, which is used to improve efficiency and accuracy. The experimental results show that the novel recognition method can not only improve the recognition accuracy, but also meet the real-time requirements of online detection.