图像分割与分类相结合的PolSAR图像机场跑道区域检测  

Airport Runway Detection Based on Figure Segmentation and Classification in PolSAR Image

作　　者：韩萍[1] 刘亚芳 韩宾宾 程争 HAN Ping;LIU Yafang;HAN Binbin;CHENG Zheng(Tianjin Key Lab for Advanced Signal Processing,Civil Aviation University of China,Tianjin 300300,China;Basic Experiment Center,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学智能信号与图像处理天津市重点实验室,天津300300 [2]中国民航大学基础实验中心,天津300300

出　　处：《信号处理》2021年第11期2084-2096,共13页Journal of Signal Processing

基　　金：国家重点研发计划(2016YFB0502405);中央高校基本科研业务费专项资金项目(3122019110,3122019046)。

摘　　要：目前全极化合成孔径雷达(Polarimetric Synthetic Aperture Radar,PolSAR)图像的机场跑道检测以基于分类的方法为主,检测效率较低,且需要更多的先验信息。针对此问题,本文提出一种基于图像分割与分类的无监督机场跑道检测方法。首先定义了图像的伪散射功率(Pseudo Scattering Power,Pspan),并结合Pspan和Tsallis熵对图像进行阈值分割,得到机场跑道感兴趣区域;然后利用密度峰值搜索和Wishart距离结合的分类器进行二分类,最后根据机场跑道的纹理和几何特征对分类结果进行进一步辨识,确定真实机场跑道区域。分别利用不同系统采集的多组实测PolSAR数据进行实验,结果表明该方法能够检测出图像中的所有机场,且外部轮廓清晰,结构完整,检测效率提高。At present,the airport runway detection in Polarimetric Synthetic Aperture Radar(PolSAR)images is mainly based on classification methods,which is inefficient and requires more prior information.To solve this problem,an unsupervised runway detection method based on image segmentation and classification is proposed.Firstly,the pseudo scattering power(Pspan)of the image is defined,and image is thresholded by using Pspan and Tsallis entropy,so as to get the region of interest.Then,the classifier combining density peak search and Wishart distance is used for binary classification.Finally,according to the texture and geometric features of airport runway,the classification results are further identified to determine the real airport runway area.Experiments were carried out by using multiple groups of measured PolSAR data collected by different systems.The results show that the proposed method can detect all airports in the image,with clear outlines and complete structure,and the detection efficiency is improved.

关 键 词：极化合成孔径雷达 图像分割 伪散射功率 密度峰值搜索 无监督分类 机场跑道检测 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]