基于HSV色彩空间的遥感图像快速云检测

Hybrid Cloud Detection Algorithm Based on HSV Color Space

对可见光波段遥感图像进行HSV色彩空间变换,结合图像的亮度特征和饱和度特征,对饱和度通道和亮度通道分别进行同态滤波和直方图均衡化处理,增强目标云区对比度;转换为RGB色彩空间之后,接着进行2%的线性拉伸,结合云的暗原色通道特性生成暗通道图,不断突出云区与非云区对比;最后进行otsu阈值分割提取云区。实验依托实际项目需求,针对原始影像波段少的特点,选取3幅压缩后全色遥感影像为实验数据,并与阈值法进行目视判别和精度统计对比。实验结果表明该方法可以快速、较高精度地检测出云区,所提出的方法对检测数据质量和波段数要求低,检测快速,检测云区较为完整准确、实用性强。

Firstly,we converted the images from RGB to HSV color space to obtain the image brightness channel and saturation channel.And then,we combined with the features of the brightness and the saturation to deal with the image with homomorphic filtering and histogram equalization respectively,to enhance the target cloud area contrast.After converting to RGB color space,the linear stretching of 2% was followed,and the dark channel diagram was generated based on the dark channel characteristics of the cloud,and the contrast between the cloud region and the non-cloud region was constantly highlighted.Finally,we performed oTSU threshold segmentation to extract the cloud region.Relying on the actual project,given the lack of original image spectrum,the experiment selected three compressed panchromatic remote sensing images as experimental data,and compared the results with threshold method by visual discrimination and accuracy comparison of statistics.The experimental results show that this method can detect the cloud area quickly and accurately.The method proposed in this paper has a low requirement for the detection data quality and band number,with a fast,complete and accurate detection of cloud area,and strong practicability.