大旋转角建筑的极化定向角估计问题和检测方法研究

Analysis of POA estimation problem and area detection method for largely-oriented structures

通过定向角补偿无法解决大旋转角建筑物的体散射估计过高的问题,需要特殊的检测和分解方法。从原理上分析了大旋转角建筑的定向角估计值偏小和精度较差的原因,指出无法根据定向角值检测大旋转角建筑。针对现有几种大旋转角建筑检测方法无法有效区分大旋转角建筑和山丘的问题,提出在已有算法的基础上引入极化散射角,用阈值来区分的解决方法。利用实测数据分析得到极化散射角的阈值,经统计,大旋转角建筑和山丘的分类正确率可达到95%。最后,用C频段的RADARSAT-2数据验证了所提方法的有效性。

Orientation Angle Compensation is unable to solve the overestimated volume scattering problem of the largelyoriented structures in Pol SAR images,thus special detection and decomposition algorithms are needed. The causation of the underestimation and the poor precision of OAs are analyzed,which explains the inefficacy of detection using OA values. The existing detection methods for largely-oriented structures cannot distinguish largely-oriented structures from the mountainous areas,in order to solve this problem,a method using alpha angle and a proper threshold is proposed. The threshold is obtained by proper analyses of measured data sets,statistics indicate that the classification accuracy of largely-oriented structures and mountainous areas can reach 95%. The validity of the proposed method is demonstrated using the RADARSAT-2 San Francisco pol SAR data set.