利用成像匹配与成分分离的航空图像云检测

Aerial image cloud detection using imaging matching and component separation

研究了遥感图像云检测,提出了基于几何成像匹配与成分分离的航空图像云检测算法。由于传统的云检测算法未考虑云具有半透明性的特点,直接提取的云纹理包含有多余的下垫面纹理。该算法依据线性光谱混合模型,将一幅遥感图像看作是由下垫面与云的光谱线性地构成的,考虑像素间的局部平滑,进行云成分分离,采用局部二值模型(LBP)特征描述云的纹理,进而通过构建支持向量机(SVM)分类器进行云检测。对比实验结果表明,本文方法对于航空图像云检测具有一定的效果,对于薄云区域以及边缘区域也能有效检测。

Cloud detection of remote sensing images is studied,and an aerial imaging cloud detection algorithm using geometric imaging matching and component separation is proposed.Directly extracted cloud textures by the existing cloud detection algorithm contain redundant underlying surface textures duo to the ignoring of translucency of the clouds.The proposed algorithm regards that a remote sensing image is composed of the spectrum of the underlying surface and the clouds linearly,according to linear spectral mixture model.Considering the local smoothing between pixels,the cloud component separation is processed.After that,local binary pattern(LBP)features are used to describe the textures of the clouds,and the clouds are detected by constructing support vector machine(SVM)classifier.The experimental results show that this method is effective for aerial image cloud detection,as well as the thin cloud area and the edge region.