基于多类支持向量机的板形识别方法

提出一种基于多类支持向量机理论的板形识别分类器，通过对冷轧工序中板形仪测得的数据进行预处理，获取所需样本数据。采用“一对多”方法训练多类支持向量机分类器，最后用测试样本对训练出的分类器进行性能测试。仿真结果表明该方法在处理小样本数据时识别率非常高，泛化能力更强，为板形识别提供了新的研究方法。

Research on flatness recognition based on M-SVMs

In this paper, a Multi-Classification SVMs classifier in terms of the theory of SVM is presented and which can tell the various properties of panel surface. The sample data is obtained by preprocessing the data which is measured through the flatness detector in the cold-rolled operation. Using the supervised method of ＂one-class-against-the-rest＂ to train Multi-Classification SVMs classifier. Finally, testing the performance of classifier by test data. The simulation results show that the proposed method performs high recognition rate in processing little sample data and has good ability of generalization. To the flatness recognition, this is a new researched method.