针对全色CCD光学遥感卫星成像平台,提出了一种基于多属性融合的高分辨率图像云检测方法。首先根据云的物理特性和成像特性展开多属性分析,并提出了基于可分离度的特征选择准则来进行特征参量提取;然后针对两类样本的分布结构差异,结合LDA线性特征压缩算法,削减特征空间的维度;最后利用SVM分类器完成云检测。实验结果表明,该方法适用于高分辨率全色遥感图像的云检测,具有较高的检测概率和较低的虚警概率以及较高的算法执行效率。

For panchromatic CCD optical remote sensing satellite imaging,a cloud detection method in high-resolutionremote sensing image based on multi-attribute fusion is proposed.First,according to the physics and imaging features,the feature parameters are extracted,and a concept of separable degree is proposed for feature selection.Then,accordingto the distribution differences of the samples,the LDA algorithm is brought for reducing the dimension of the featurespace.At last,a SVM classifier is used for cloud detection.The experimental results show that the proposed method issuitable for high-resolution panchromatic remote sensing image with different imaging platform,and it has a high detec-tion probability and a low false alarm probability,and a high efficiency.