论文部分内容阅读
利用全极化SAR影像进行城市地区地物分类和目标识别时,极化方位角偏移、散射类型复杂多样以及弱后向散射地物难以区分等问题影响了图像分类和目标识别的准确性。针对这些问题,提出一种利用目标散射矢量模型TSVM(Target Scattering Vector Model)生成的旋转不变极化参数组合提取城市区域典型人工地物(建筑、道路、广场和桥梁等)的方法。首先,采用目标散射矢量模型分解和Relief特征选择算法提取出表征目标对称性的旋转不变极化参数、表征散射机理的散射角旋转不变极化参数;然后利用支持向量数据描述SVDD(Support Vector Data Description)分类器快速提取典型人工地物。利用南京市Radarsat-2全极化SAR数据进行实验,结果表明提出的方法抑制了地形起伏的山地或建筑分布不规则城区的极化方位角效应,有效解决了全极化SAR影像中人工地物提取的难题。同时,相比光学影像利用生物物理组合指数BCI(Biophysical Composition Index)和归一化建筑指数NDBI(Normalized Difference Built-up Index)提取城市人工地物的方法,对称性极化参数和散射角参数组合能有效区分出城市建筑用地和具有相似光谱特征的裸地,其检测精度提高10%以上。
When using polarimetric SAR images to classify and target objects in urban areas, the problems such as the azimuthal polarization shift, the complex and diverse types of scattering, and the difficulty of distinguishing the features of weakly backscattered images affect the accuracy of image classification and target recognition. In order to solve these problems, this paper proposes a method for extracting typical artificial objects (buildings, roads, squares and bridges, etc.) in the urban area by using the invariant polarimetric polarization parameters generated by the Target Scattering Vector Model (TSVM). First, the invariant polarization parameters characterizing the target symmetry are extracted by using the target scattering vector model decomposition and the Relief feature selection algorithm, and the invariant polarization parameters of the scattering angle are characterized. Then the support vector data is used to describe the SVDD Data Description Classifier Quickly extract typical artifacts. Experiments using the Radarsat-2 full-polarimetric SAR data in Nanjing show that the proposed method suppresses the azimuthal polarization effect in the irregularly distributed mountainous areas or irregularly distributed urban areas, and effectively solves the problems of artifacts in the fully polarimetric SAR images Extraction of the problem. At the same time, compared with the method of extracting urban artificial features by optical image using Biophysical Composition Index (BCI) and Normalized Difference Built-up Index (NDBI), the combination of symmetry polarization parameters and scattering angle parameters Can effectively distinguish between urban construction sites and bare land with similar spectral characteristics, the detection accuracy increased by 10%.