衛星遙測影像被用來觀察與分析地表地物的變化已經有好幾十年。然而，表面變化的自動分析結果容易受到影像中雲霧的干擾。因此，雲霧偵測成為遙測領域中的一個重要議題。在過去，基於閾值的雲偵測演算法已經發展成熟並且被應用於遙測領域的各種任務上。在這些方法中，Fmask演算法針對Landsat-8衛星影像的雲辨識精確度已經達到90%以上，而SCL則是針對Sentinel-2衛星影像有很好的辨識能力。然而這些方法的識別結果仍然不夠可靠，在偵測比一般的雲更難偵測的卷雲時，準確度太低，不足以應用於自動分析任務上。近年來，隨著深度學習的快速發展，許多針對語意圖像分割的深度學習方法已經被廣泛接受並成功地應用於各種物件偵測與辨識的任務。在這個研究中，一個結合了Resnet，Atrous Convolution與FCN的架構被提出了。為了訓練與驗證我們的深度學習模型，100張標準大小為224 像素的人工標記Sentinel-2衛星影像被用於此研究。在這些圖像中，20張圖像被保留作為驗證資料集而80張圖像藉由資料擴增技術擴增為10,000張圖像後作為訓練資料。此研究比較了我們提出的方法與基於閾值的SCL方法，提出的方法在評估指標Pixel Accuracy得到接近94%。比SCL有更高的辨識雲的能力。

Satellite remote sensing imagery have been used to observe and analysis changes of earth surface and land cover for decades. However, the automatic analysis of earth surface changes is susceptible to interference from cloud and haze in the image. Therefore, cloud detection becomes an important issue in the field of remote sensing. In the past, threshold-based cloud detection algorithms have been well developed and applied in various of tasks in the field of remote sensing. Among these methods, Fmask algorithm achieves more than 90% in pixel accuracy for Landsat-8 imagery and SCL algorithm have a good cloud recognition capability for Sentinel-2 imagery. However, the detection results of these methods are not reliable enough to be applied to automated analysis tasks since to low accuracy when detect cirrus clouds which are more difficult to detect than normal clouds. In recent years, with the rapid development of deep learning, many methods for semantic image segmentation have been widely accepted and successfully applied to various object detection and identification tasks. In this study, In this study, an architecture combining Resnet, Atrous Convolution and FCN was proposed. In order to train and validate our deep learning model, 100 manually labeled Sentinel-2 images with standard size of 224 x 224 pixels are used for this study. Among these images, 20 images are preserved as validation dataset and rest of them are expanded into 10,000 images by data augmentation techniques and used as training dataset. This study compares proposed method with threshold-based method SCL. The proposed method achieves 94% in pixel accuracy.

1. LeCun, Y.; Boser, B.; Denker, J.S.; Henderson, D.; Howard, R.E.; Hubbard, W.; Jackel, L.D. Backpropagation applied to handwritten zip code recognition. Neural Computation 1989, 1, 541-551.
2. Rumelhart, D.E.; Hinton, G.E.; Williams, R.J. Learning representations by back-propagating errors. Nature 1986, 323, 533.
3. Lecun, Y.; Bottou, L.; Bengio, Y.; Haffner, P. Gradient-based learning applied to document recognition. Proceedings of the IEEE 1998, 86, 2278-2324.
4. Deng, J.; Dong, W.; Socher, R.; Li, L.; Kai, L.; Li, F.-F. In Imagenet: A large-scale hierarchical image database, 2009 IEEE Conference on Computer Vision and Pattern Recognition, 20-25 June 2009, 2009; pp 248-255.
5. Krizhevsky, A.; Sutskever, I.; Hinton, G.E. Imagenet classification with deep convolutional neural networks. In Proceedings of the 25th International Conference on Neural Information Processing Systems - Volume 1, Curran Associates Inc.: Lake Tahoe, Nevada, 2012; pp 1097-1105.
6. Long, J.; Shelhamer, E.; Darrell, T. Fully convolutional networks for semantic segmentation. In ArXiv e-prints, 2014; Vol. 1411.
7. Ronneberger, O.; Fischer, P.; Brox, T. U-net: Convolutional networks for biomedical image segmentation. In ArXiv e-prints, 2015; Vol. 1505.
8. Chen, L.-C.; Papandreou, G.; Kokkinos, I.; Murphy, K.; Yuille, A.L. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. In ArXiv e-prints, 2016; Vol. 1606.
9. Xie, F.; Shi, M.; Shi, Z.; Yin, J.; Zhao, D. Multilevel cloud detection in remote sensing images based on deep learning. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 2017, 10, 3631-3640.
10. Zhan, Y.; Wang, J.; Shi, J.; Cheng, G.; Yao, L.; Sun, W. Distinguishing cloud and snow in satellite images via deep convolutional network. IEEE Geoscience and Remote Sensing Letters 2017, 14, 1785-1789.
11. Drönner, J.; Korfhage, N.; Egli, S.; Mühling, M.; Thies, B.; Bendix, J.; Freisleben, B.; Seeger, B. Fast cloud segmentation using convolutional neural networks. Remote Sensing 2018, 10, 1782.