摘要
合成孔径雷达(Synthetic Aperture Radar,SAR)图像自动目标识别的前提条件之一是能够准确地提取感兴趣区域(Region of interest,ROI),因此能够获取ROI中心的聚类算法是SAR图像处理的关键算法之一。为了尽可能降低检测图像中的虚警以及减少聚类及相应的鉴别算法的计算量,本文提出一种基于先验信息的网格聚类算法,该方法首先通过目标和杂波的形状统计信息估计网格聚类参数,然后利用其对检测图像进行网格划分,并引入目标的占空比特征去除杂波,最后通过粗提取和精提取两种方法计算得到聚类中心。仿真和实测数据处理结果表明,该算法能够对检测目标进行有效聚类并去除大部分杂波,同时极大地减少了鉴别的计算量,且简化了传统ROI中心提取流程。
For synthetic aperture radar (SAR) images, the key to successful automatic target recognition is accurate extraction of regions of interests (ROI), and as a consequence, the clustering algorithm is the key to ROI extraction. A novel approach based on grid clustering is proposed in this paper, which can significantly reduce the false alarm rate and the computational complexity. Firstly, the statistics of geometrical features are gathered to determine the parameters of the algorithm; then, the whole duty ratio map is generated by computing the duty ratio in each grid and the oversized and undersized clutters is discarded according to the corresponding duty ratios ; at last, the ROI centers are estimated from the duty ratio map. The feasibility and effectiveness of our proposed approach are demonstrated by simulation and real experimental results.
出处
《信号处理》
CSCD
北大核心
2012年第11期1565-1574,共10页
Journal of Signal Processing
基金
国家自然科学基金项目(61271441
60972121)
全国优秀博士学位论文作者专项资金资助项目(201046)
新世纪优秀人才支持计划资助项目(NCET-10-0895)
国防科技大学科研计划项目(CJ12-04-02)
关键词
先验信息
网格聚类
时间复杂度
Prior Information
Grid Clustering
Time Complexity