一种改进谱聚类的机体损伤图像过渡区提取方法

发布时间：2018-03-31 13:10

本文选题：机体损伤过渡区提取　切入点：机体损伤区域划分　出处：《计算机辅助设计与图形学学报》2016年10期

【摘要】：为了利用机体损伤区域图像支持飞机智能维修,提出一种改进谱聚类的机体损伤过渡区提取方法.在分析机体损伤邻接区域特点的基础上,以多维灰度熵值作为相似性度量构建样本点谱图实现了谱图权重的降维运算;并以统计分组中首个极小值点所对应的权重为临界进行聚类;通过对谱聚类方法中的关键环节特征值计算、特征向量选取进行改进,避免了因特征向量的取舍而造成的信息丢失.最后应用机体损伤图像实例验证了该方法的有效性.
[Abstract]:In order to support the intelligent maintenance of aircraft using the image of airframe damage region, an improved spectral clustering method for extracting the transition region of airframe damage is proposed.On the basis of analyzing the characteristics of the adjacent area of organism damage, the dimensionality reduction of the weight of the spectrum is realized by constructing the sample point spectrum with multidimensional gray entropy as similarity measure.The weight corresponding to the first minimum point in the statistical grouping is taken as the critical value for clustering, and the eigenvector selection is improved by calculating the eigenvalues of the key links in the spectral clustering method.The information loss caused by the choice of feature vectors is avoided.Finally, the effectiveness of the method is verified by an example of body damage image.
【作者单位】：中国民航大学航空工程学院;
【基金】：航空科学基金(20151067003)
【分类号】：V267;TP391.41

【相似文献】