基于SVM分类的边缘提取算法

发布时间：2018-08-31 18:14

【摘要】：通过分析同类数据点在空间中的几何形态,从数据点集所构成几何形态的凹凸性着手,提出边界提取算法并对高维数据进行分类。针对现实生活中的高维数据,利用局部线性嵌入将数据进行降维处理,得到低维特征数据。在此基础上,对于单分类数据集,用数据集表面的点的近邻样本与过该点的切平面之间的关系寻找边界点;对于多分类数据集,利用贝叶斯后验概率来寻找边界重复的点,以此更快达到提取边界点的目的。由此可以粗略筛选出边界点。为去除不重要的边界点,降低分类误差,通过构造最优超平面和支持向量机对边界点赋予权重,并设置阈值去除不重要的边界点,由此达到用较少的边界点准确分类数据的目的。通过100个测试样本进行分类测试并计算其分类准确率,验证了此分类方法的可行性。
[Abstract]:By analyzing the geometric shape of the same data points in space, starting with the concave and convex nature of the geometric form formed by the data points set, a boundary extraction algorithm is proposed and the high-dimensional data are classified. For the high-dimensional data in real life, the low-dimensional feature data is obtained by using local linear embedding to reduce the dimension of the data. On this basis, for a single classification dataset, the relationship between the nearest neighbor sample of a point on the surface of the dataset and the tangent plane passing through the data set is used to find the boundary point, and for the multi-classification data set, the Bayesian posteriori probability is used to find the point of boundary repetition. In this way, the boundary point can be extracted more quickly. The boundary points can be roughly screened. In order to remove the unimportant boundary points and reduce the classification errors, the optimal hyperplane and support vector machine are constructed to give weight to the boundary points, and the threshold is set to remove the unimportant boundary points. Thus, the purpose of classifying data accurately with fewer boundary points is achieved. The feasibility of this classification method is verified by 100 test samples and its classification accuracy is calculated.
【作者单位】：成都理工大学管理科学学院;
【基金】：四川省社科规划项目(SC16TJ004)
【分类号】：TP18;TP311.13

【相似文献】