基于改进模糊支持向量机的磁记忆检测缺陷识别

针对磁记忆检测信号弱、缺陷区域无法有效识别的问题，提出了一种改进的模糊支持向量机(FSVM)，并将其应用于磁记忆检测缺陷的识别。改进的FSVM一方面在传统确定模糊隶属度函数方法的基础上，通过构造k近邻离散度，减弱孤立点或噪声样本对分类的影响；另一方面通过对样本特征值进行加权处理，消弱冗余特征或弱特征对识别的影响。将改进FSVM应用于磁记忆检测缺陷识别。实验结果表明：该方法可以有效识别不同危险区域的缺陷信号，具有较好的鲁棒性和分类能力，是一种有效的磁记忆检测缺陷识别方法。

Defect Recognition of Metal Magnetic Memory Testing Based on Improved Fuzzy Support Vector Machine

Since the metal magnetic memory (MMM) signal is weak and the defects can’t be recognized effectively, a novel fuzzy support vector machine(FSVM) is proposed. In order to reduce the influence of isolation point and noise on classification accuracy, the?k?nearest neighbor dispersion is constructed based on the traditional determination method of the fuzzy membership. Besides, the feature weighted degree of each feature is calculated to reduce the influence of redundant and weak features on classification accuracy. And then the proposed approach is applied to recognize MMM signals of different areas, the experimental results show that the proposed approach can recognize this MMM signals effectively，it is more robust and has the better performance of recognition. The proposed FSVM approach is a feasible recognition algorithm for MMM signals of different areas.