Classified Linear Regression Based Landsat Image Cloud Removal Method
-
摘要: 首先,对参考影像和待去云影像上的云覆盖区域进行检测和掩膜;然后,对掩膜后的参考影像进行ISODATA聚类,并建立各个类别参考影像到待去云影像灰度值的线性回归方程;再对待去云影像上的云覆盖区域,依据参考影像上的灰度值进行最小距离方法分类,划分到聚类形成的各个类别之中;最后,依据各个类别回归方程进行灰度值预测。实验结果表明,所提方法能够进行云区的检测和去除,预测精度相比传统方法有较大提高。Abstract: An approach for cloud removal based on linear regression after image classification is proposed in this article.First of all,the clouds in a remote sensing image and its referenced data to be processed are detected,from which two cloud masks are built.Then,an ISODATA classification is applied to the referenced image with the cloud mask.Next,the masked part of the contaminated image is classified with the existing clusters of the referenced data using the minimum distance method.Last,the digital numbers of the cloudy areas of the contaminated image are replaced with by the prediction value of the referenced data calculated by the linear relationships determined between clusters on the referenced image and the corresponding contaminates done according to the pixel location.This algorithm is programmed to automatically detect and remove the clouds areas in Landsat images.The accuracy of cloud detection and the prediction of original values of the cloud cover are evaluated.Results show that the proposed method is effective.
-
Keywords:
- image mosaic /
- cloud removal /
- cloud detection /
- classified linear regression
-
-
[1] 王伟宋卫国,刘士兴,张永明,郑红阳,田伟,. Kmeans聚类与多光谱阈值相结合的MODIS云检测算法[J]. 光谱学与光谱分析. 2011(04)[2] 陈奋闫冬梅,赵忠明,. 基于无抽样小波的遥感影像薄云检测与去除[J]. 武汉大学学报(信息科学版). 2007(01)[3] 吴传庆王桥,杨志峰,. 基于混合像元分解的水体遥感图像去云法[J]. 遥感学报. 2006(02)[4] 马建文,顾行发,冯春,郭建宁,. CBERS-02卫星图像薄云的去除方法研究[J]. 中国科学E辑:信息科学. 2005(S1)
计量
- 文章访问数: 2182
- HTML全文浏览量: 91
- PDF下载量: 881
下载:
