基于拒识的误差编码车辆分类

基于支持向量机的误差编码分类已经成功应用到多类分类问题中，但通常是以错误率来衡量误差编码分类器的优劣，具有较大的局限性。在实际应用中，错分结果往往会带来一定的风险代价，为了在一定程度上减少这种代价，本文提出了一种带有拒识功能的误差编码分类方法。文中利用纠错编码把多类分类问题转换成多个两类分类，再将拒识的方法应用到每个分类器中，最后结合编码矩阵投票表决得到分类结果，从而提高分类可靠性。

Error Correcting Output Coding Vehicle Classification with Rejection

Support vector machine (SVM) based on error correcting output coding classification has been successfully applied to the multi-class classification problem. Conventionally, error rate is employed to measure the performance of SVM which brought a big limitation. To overcome this limitation, we propose a vehicle classification based on error correcting output coding with rejection. Via this approach, the multi-class classification problem is converted into binary classification problem. By carrying out inner rejection the general reliability of SVM in classification problem is enhanced.