Abstract: With the analysis of the features of infrared image, an infrared weak target segmentation method based on mountain clustering was presented. First, the original image was projected to x and y coordination to construct dataset and used as sample to cluster. Then based on the datasets, the improved mountain clustering algorithm was implemented, and the clustering centers were employed as the seeds in region growth, and at last the final segmentation was acquired. Experimental results show that the proposed method outperforms the traditional approach and it can provide satisfactory segmentation result.

Key words: infrared image, target segmentation, mountain clustering, projection, region growing

摘要： 在分析目标红外图像特征的基础上，利用山峰聚类对密度大样本的快速聚类特点，提出了一种基于山峰聚类的弱目标区域生长分割方法。该方法首先将二维图像在x和y方向分别进行投影并以此为样本构造数据网格，再采用改进的山峰聚类算法进行聚类，将求出的聚类中心作为种子点进行区域生长得到分割结果。实验结果表明，与传统方法比较，所提方法能自动选取种子点，投影后数据作为样本可有效提高运算速度，并能取得良好的分割效果。

关键词: 红外图像, 目标分割, 山峰聚类, 投影, 区域生长