引入深度优先策略的图像边缘聚类算法仿真

Image Edge Clustering Algorithm Simulation Introduce Depth-first Strategy

为了实现对提取边界后剩余数据对象的聚类,提出一种由图像边缘出发进行聚类的算法。该算法首先采用深度优先搜索的策略将已知的边界对象进行分类,并计算各边界曲线的最小外包矩形区域;然后运用夹角和法去除内边界类;最后依据近邻原则对每一个核心对象进行归类。实验结果表明,对于含有噪声、密度均匀的数据集,算法可以识别出各种形状的聚类,且聚类质量和时间性能较好。

In order to cluster the remaining data objects after extraction of the border, a kind of clustering algorithm based on the minimum bounding rectangle of border is proposed. Firstly, it classify the known boundary object in accordance with the depth-first search, and calculate the minimum bounding rectangle area of every boundary curve. And then exclude from the inner boundary of cluster using the method of the sum of the angle. Finally, cluster the core object cluster by neighbor principle. The experimental results show that, for containing noise, density data set, the algorithm can identify various shapes cluster, and clustering quality and time performance is better.