骨架提取算法的研究及在异形纤维中的应用

A Study of Skeleton Determination and the Application in Profiled Fibers

为了对异形纤维进行分类识别,需要对主分枝进行计数,提取纤维骨架是重要的预处理方法.提出利用物体边缘点和最小覆盖集的骨架生成算法,提取异形纤维的骨架信息.通过计算二值图像物体的局部中心点集,结合物体边缘点建立相关矩阵,获得最小覆盖集,然后利用爬山算法结合方向导数生成连续骨架.最后,对细小分枝进行修剪,保留纤维主分枝.实验证明,本算法计算的骨架保存了物体拓扑信息,并且对物体边缘噪声具有较强的抗干扰能力.

The determination of profiled fiber＇s skeleton is an important pre-processing method for profiled fiber＇s identifying because the number of branches is necessary to be calculated. A skeleton determination algorithm was proposed, in which the minimum cover set can be obtained firstly through the calculation of the correlated matrix with the local central point set and the edge information, and then the uphill algorithm was used to produce a continued skeleton, finally the main branch of fiber was retained after pruning the small branch. The fiber＇s topology information is well preserved in the algorithm, which has good noise performance as well as low computational complexity showed by experiments.