A Cloud Detection Algorithm for Remote Sensing Images Using Fully Convolutional Neural Networks

Abstract

This paper presents a deep-learning based framework for addressing the problem of accurate cloud detection in remote sensing images. This framework benefits from a Fully Convolutional Neural Network (FCN), which is capable of pixellevel labeling of cloud regions in a Landsat 8 image. Also, a gradient-based identification approach is proposed to identify and exclude regions of snow/ice in the ground truths of the training set. We show that using the hybrid of the two methods (threshold-based and deep-learning) improves the performance of the cloud identification process without the need to manually correct automatically generated ground truths. In average the Jaccard index and recall measure are improved by 4.36% and 3.62%, respectively.

Cloud-Net: An End-To-End Cloud Detection Algorithm for Landsat 8 Imagery

“…Jaccard Index = T P T P + F N + F P Here TP, TN, FP, and FN are the total number of true positive, true negative, false positive, and false negative pixels, respectively. The Jaccard Index or intersection over union is a widely accepted metric for measuring the performance of many image segmentation algorithms [13]. Table 1 represents the quantitative results of the proposed method over 20 test images of the 38-Cloud dataset.…”

“…Therefore, the numerical results of this experiment are fair to be compared with that of Cloud-Net. Please note that according to [13], FCN's overall accuracy is 88.30%. Also, the proposed method exceeds Fmask's performance, which is a widely used algorithm for cloud detection.…”

“…In recent years, researchers have developed many cloud detection methods. These methods can be divided into three main categories: threshold-based [6,7,8,9], handcrafted [10,11], and deep learning-based [12,13] approaches. Zhu et al in [7] introduced the Function of mask (Fmask) algorithm.…”

“…The effectiveness of U-Net has been proven in many other computer vision applications [16,17]. Authors in [13] used a FCN to segment the cloud regions of Landsat 8 images. This network was a very deep version of U-Net and was trained by images and their automatically generated Ground Truths (GTs).…”

“…It does not require any complicated pre-processing step such as super-pixel segmentation. In addition, unlike [13], Cloud-Net effectively utilizes the extracted features from its sophisticated convolution blocks to recover more accurate cloud masks. As a result, it delivers superior performance for cloud detection.…”