国产高分多光谱数据的自动云检测

Cloud Detection Algorithm for Domestic High-Resolution Multispectral Image Data

针对国产高分多光谱影像质量判定的需求,提出阈值法与聚类分析相结合的云检测算法。分析大量高分多光谱影像中云区的特征,计算云区值域的统计区间。根据统计结果,首先识别出影像中云区的种子区域和地物的种子区域。利用相似性聚类分析,进一步细化未识别区域,检测出影像中完整的云区。该方法在大量高分多光谱影像数据上进行实验,均得到较好结果。

In order to conduct quality judgement of domestic high-resolution image, the cloud-detection algorithm is proposed, which combines the threshold value method with the cluster analysis. By studying the features of clouds in many high resolution multispectral images, the statistical area of cloud value range is calculated. According to the statistical result, the algorithm can be used to identify the seed regions of cloud and object. The unlabeled region is further refined and the whole cloud region is detected by using similarity cluster analysis. Some experiments are conducted on a large number of high-resolution image data are conducted, and the results show that this algorithm has high performance.