基于图割模型的高分全色影像自动云检测

Automatic Cloud Detection for High Resolution Panchromatic Images Based on Graph Cuts Model

针对高分全色卫星影像场景复杂、波段单一、波谱范围有限的特点,提出了一种基于图割模型的自动云检测方法。该方法首先通过亮度双阈值和Gabor纹理特征将影像分为云区、非云区和不确定区域;再计算像素点的能量项和平滑项;然后通过图割模型确定不确定区域像素点的类别,从而检测出精确的云区。结果表明,该方法能有效检测高分全色影像上的云区范围,且无需人工干预,可满足自动化生产需求。

High resolution panchromatic images have complex scenes,single band and limited spectral range.Focusing on those features,this paper put forward an automatic cloud detection method based on graph cuts model.Firstly,this method segmented an image into cloudy areas,none cloudy areas and possible areas by gray bi-threshold segmentation and Gabor feature.And then,the method calculated the data term and smooth term.Finally,the accurate cloudy areas were detected via graph cuts model.The results show that this method can detect the cloudy areas in high resolution panchromatic images,and the procedure is automatic,which is suitable for the requirements for production automation.