摘要： 提出一种有效的遥感图像阵列目标识别方法，不仅利用空间属性特征，而且通过设计Gabor滤波器组提取子目标的非空间属性特征。在对子目标进行聚类时，根据图划分理论给出一种迭代提取样本中相似样本群的方法，实现子目标群的提取。实验结果表明，通过加入非空间属性特征和采用迭代提取样本中相似样本群的方法，可以有效地排除大量虚假子目标并精确定位阵列目标。

关键词: 阵列目标, Gabor滤波器, 特征聚类

Abstract: An effective algorithm for array targets recognition in remote sensing images is proposed. It not only utilizes spatial features, but also extracts Gabor texture as non-spatial features for feature clustering. An iterative analogous array targets extraction process is used based on graph cuting theory, aiming at locating the targets array. Experiments show that by adding the non-spatial features and using the iterative analogous array targets extraction process, the proposed algorithm is able to eliminate a large number of false alarms and locate the array target accurately.

Key words: array targets, Gabor filter, feature clustering

中图分类号: 

• TP18