一种改进的支持向量机及其在图像分割中的应用

多类支持向量机在实际应用领域是一个非常重要的问题。广泛应用的多类SVM方法包括：一对一、一对多和DAG等。众多实验表明一对一方法通常具有较高分类准确率,但传统一对一方法测试时间较长限制了其在大数据量识别任务中的应用。针对一对一支持向量机方法进行了改进,提出了一种改进的支持向量机,并采用其对多目标图像进行了分割研究。实验结果表明,支持向量机方法是一种很有前景的图像分割技术。

An Improved Support Vector Machine and Its Application in Image Segmentation

Multi-class support vector machine has a very important issue in the field of practical application.The widely used multi-class SVM methods include： one-against-one,one-against-all and DAG.Many experiments show that one-against-one method usually has high classification accuracy.But the traditional one-against-one method takes a longer time;it limits its application in the amount of data in the recognition task.An improved support vector machine was suggested which is based on one-against-one support vector machine and used to image segmentation.Experiments show that support vector machine approach is a promising image segmentation technique.