针对传统支持向量机(SVM)对图像中含有的噪声或野值样本的不敏感性问题,提出了一种基于双向隶属度的模糊支持向量机(FSVM)的方法。该方法通过计算样本类中样本与其所属类别和另一类别的中心点之间的距离,得出样本对每一类的隶属度,通过对样本隶属度的分析实现对样本点的分类。将该方法应用于对肺结节良恶性分类试验中,其参数反演结果表明,即使在噪声存在的情况下该方法也能得到较好的分类结果,而且克服了传统方法过拟合的缺点,从而也验证了该方法较强的抗噪能力和较好的分类能力。 Aiming at the insensitivity of traditional support vector machines (SVM) to noise or outliers in images, a fuzzy support vector machine (FSVM) method based on two-way membership is proposed. The method obtains the membership degree of each sample by calculating the distance between the samples in the sample category and the center point of another category and the classification of the sample points by analyzing the sample membership degree. The method was applied to the benign and malignant pulmonary nodule classification tests, the inversion results show that the method can get good classification results even in the presence of noise, and overcome the shortcomings of over-fitting traditional methods , Which also verified the method of strong anti-noise ability and better classification ability.