基于多级分类的PolSAR图像机场跑道检测

发布时间：2018-03-20 06:25

本文选题：极化合成孔径雷达图像　切入点：复杂场景　出处：《系统工程与电子技术》2014年05期 　论文类型：期刊论文

【摘要】：提出一种复杂场景极化合成孔径雷达(polarimetric synthetic aperture radar,PolSAR)图像机场跑道多级分类检测方法。首先利用先验信息进行h/q第一级分类,得到图像中各类训练样本模板;然后利用PolSAR图像极化相干矩阵的统计特性进行第二级分类;再根据跑道的弱回波特性,利用极化总功率检测器完成第三级分类,提取图像中疑似机场跑道区域;最后根据机场跑道的尺寸及结构特征进行判别,确定机场跑道区域。采用美国UAVSAR机载系统获取的多组实测数据对算法进行验证,并与现有的两种方法进行比较,结果表明,本文算法能有效地检测出跑道,并保持跑道结构完整、轮廓清晰且虚警率低。
[Abstract]:In this paper, a multi-level classification method for airport runway in polarimetric synthetic aperture radar (PolSAR) images of complex scene is proposed. Firstly, the hrq first stage classification is carried out using prior information, and various training sample templates in the image are obtained. Then the second stage is classified by using the statistical characteristics of the polarimetric coherence matrix of the PolSAR image, and the third level classification is completed by using the polarimetric total power detector according to the weak echo characteristics of the runway, and the suspected airport runway area is extracted from the image. Finally, according to the size and structural characteristics of the airport runway, the region of the runway is determined. The algorithm is verified by using a number of measured data obtained by the UAVSAR airborne system in the United States, and compared with the existing two methods. The results show that, The algorithm can effectively detect the runway and keep the runway structure intact, the contour is clear and the false alarm rate is low.
【作者单位】：中国民航大学天津市智能信号与图像处理重点实验室;东方通用航空有限责任公司;
【基金】：国家自然科学基金重点项目(61231017) 中央高校基本科研业务费专项基金(3122014C004) 中国民航大学科研基金(2012KYE03)资助课题
【分类号】：TN957.52

【参考文献】