A Cloud Detection Method for MODIS Images Based on the Dataset from Radiative Transfer Simulations

Accurate cloud detection in high-resolution remote sensing imagery by weakly supervised deep learning

Cloud contamination is a critical source of errors in the data assimilation of hyperspectral infrared radiance (IR). Therefore, it is necessary to filter out cloudy observations. In this study, we review and summarize the principles and research progress of cloud detection methods for the hyperspectral IR in the past two decades. Based on the impact of IR data utilization on cloud detection results, cloud detection methods are categorized into five types, namely clear field-of-view (FOV) detection, clear channel detection, three-dimensional cloud detection, cloud-clearing and deep learning methods. Clear FOV methods and clear channel methods aim to identify the purely clear FOVs and spectral channels that are not affected by clouds, respectively. Cloud-clearing methods are used to reconstruct clear-column radiance for cloudy observations. Deep learning cloud detection methods can quickly learn the mapping relationship between infrared hyperspectral radiation characteristics and FOV cloud distribution from a large amount of infrared radiative information with known FOV cloud labels. In this paper, we discuss and provide an outlook on the key issues in current hyperspectral IR cloud detection. Specifically, we analyze and summarize the factors affecting cloud detection, such as surface background information, vertical cloud distribution, hyperspectral IR channel selection, improvements in cloud detection algorithms and model applicability. The results indicate the use of deep learning methods offer advantages in detection accuracy and algorithm efficiency of hyperspectral IR cloud detection.

Accurate cloud detection is a crucial step in analyzing optical satellite imagery.The increasing application of small satellites equipped only with Visible and Near-Infrared (VNIR) sensors necessitates the evaluation of cloud detection under limited spectral bands.This study focuses on the intercomparison of three cloud detection methodologies-a custom threshold-based method, a machine learning model (Extreme Gradient Boosting; XGBoost), and a deep learning model (U-Net)-using only five VNIR bands from the CloudSEN12+ dataset.As a result, the U-Net model demonstrated higher performance than other approaches, achieving an Overall Accuracy (OA) of 85.6% and an Intersection over Union (IoU) of 65.3%.In particular, the U-Net significantly outperformed others in detecting cloud shadows, achieving an accuracy of 77.4%, nearly double that of XGBoost (46.0%).In spatial analysis, the threshold method misclassified bright surfaces as clouds, and XGBoost generated salt-and-pepper noise at the pixel level.The U-Net accurately detected the complex boundaries of clouds and shadows.Our results indicate that deep learning approaches that leverage spatial information are highly effective for cloud detection with limited spectral bands.This study provides a quantitative baseline to inform the development of operational cloud-detection algorithms for VNIR-only satellite missions.

Deep learning based cloud detection for medium and high resolution remote sensing images of different sensors

With strong self-learning and data analysis capabilities, deep learning is essential in cloud detection. However, many high-quality samples are the key to deep learning cloud detection methods. Different satellite image cloud detection techniques need to select adequate representative and high-quality samples and make corresponding cloud masks, which not only requires professional knowledge but also consumes a lot of workforce and time. To improve the generalization ability of the deep learning cloud detection model and quickly apply it to the cloud detection of different satellite images, this paper proposes a deep learning cloud detection method based on spectral assimilation for multiple types of satellite images (SAUNetCD). Under the condition of using fewer deep learning data samples, the deep learning model is used to achieve automatic cloud detection of multiple satellite data. Taking Landsat 8 OLI, Landsat 9 OLI, GF-1 WFV, and Sentinel 2A as examples, this paper selects Landsat 8 OLI data as the source data of cloud detection. The experimental results show that the spectral assimilation method improves the generalization ability of the deep learning cloud detection model and improves the cloud detection accuracy by nearly 20%. It realizes the fast cloud detection application of different satellite images by deep learning. It provides an effective way for cloud detection of multiple types of satellite images.

Cloud detection is an important step in remote sensing image processing and a prerequisite for subsequent analysis and interpretation of remote sensing images. Traditional cloud detection methods are difficult to accurately detect clouds and snow with very similar features such as color and texture. In this paper, the features of cloud and snow in remote sensing images are deeply extracted, and an accurate cloud and snow detection method is proposed based on the advantages of Unet3+ network in feature fusion. Firstly, color space conversion is performed on remote sensing images, RGB images and HIS images are used as input of Unet3+ network. Resnet 50 is used to replace the Unet3+ feature extraction network to extract remote sensing image features at a deeper level, and add the Convolutional Block Attention Module in Resnet50 to improve the network’s attention to cloud and snow. Finally, the weighted cross entropy loss is constructed to solve the problem of unbalanced sample number caused by high proportion of background area in the image. The results show that the proposed method has strong adaptability and moderate computation. The mPA value, mIoU value and mPrecision value can reach 92.76%, 81.74% and 86.49%, respectively. Compared with other algorithms, the proposed method can better eliminate all kinds of interference information in remote sensing images of different landforms and accurately detect cloud and snow in images.

Cloud coverage hinders the effective range of Earth observation by optical remote sensing satellites. Rapid and accurate cloud detection is an important step in the product generation process of remote sensing applications. Given the lack of suitable and high-quality cloud detection models in the Google Earth Engine cloud platform, this study takes tropical cloudy Sri Lanka as the study area and constructs a Sentinel-2 image cloud detection model coupled with support vector machines and Cloud-Score algorithm. Through experiments, the cloud detection accuracy of this method was compared to the QA60 method, Cloud-Score algorithm, and Function of mask (Fmask) from the point of view of visual interpretation and quantitative analysis. Compared with the other three cloud detection methods, the cloud detection model proposed in this study has the highest overall accuracy, reaching 98.21%, with an extremely low omission and commission errors. The model can accurately identify the cloud boundary and meet the cloud detection pre-processing requirements of Sentinel-2 remote sensing products.

Detecting and eliminating clouds is a crucial step in remote sensing image (RSI) preprocessing. The removal of clouds can significantly enhance the performance of subsequent remote sensing applications. Existing deep learning (DL)-based cloud detection methods extract semantic information to improve feature representation and, subsequently, detection performance. However, these methods do not fully utilize the potential of context semantic information. Besides, to capture semantics from large receptive fields, they employ convolution operators with large kernel sizes, which results in high computational costs. Thus, these computationally heavy models are not suitable for resource-limited devices, particularly satellites. To address this issue, we propose a cloud detection model, LGCNet. This model efficiently extracts both local and global contextual information, fully utilizing semantics while reducing resource usage. LGCNet is built on an encoder-decoder structure. Specifically, the encoder extracts local scale-aware semantics through proposed local semantic blocks (LSBs), which are then skip-connected to the decoder. This approach provides adaptive and diverse local contextual information. On the top of the encoder, the high-level global semantics are captured via the proposed global feature TransBlock (GFTB). A variety of extracted semantics ensure improved detection performance. We evaluate the proposed method using two public datasets: LandSat8 and Moderate-Resolution Imaging Spectroradiometer (MODIS). We conducted experiments on both a server and an edge computing device. Our extensive experiments revealed that LGCNet outperforms other lightweight cloud detection and semantic segmentation methods in terms of performance and computational load.

Clouds in Sentinel-2 images seriously affect its usage in various fields, such as agricultural production and environmental monitoring. Although cloud masks have been provided in Sentinel-2 products, the stability of cloud detection accuracy may be affected by different types of underlying surfaces. This study presents a novel cloud detection method for Sentinel-2 images based on segmentation prior and multiple features. In the presented method, spectral, texture, and exponential features are extracted to enhance the difference between clouds and underlying surfaces. Meanwhile, segmentation results are regarded as priors to constrain pre-classification results to improve the edge accuracy of cloud detection and to obtain detection results with low false alarm rate and low omission rate. Experiments on the Sentinel-2 image dataset show that the presented method achieves good and stable cloud detection results, and the accuracy of cloud detection for six underlying surface types (impervious areas, water, croplands, bare lands, snow & ice, and forest) are above 0.93. These findings demonstrate that the presented method has the potential to effectively improve the stability of cloud detection accuracy while reducing the requirement for the number of samples.

Cloud detection is a key step in optical remote sensing image processing, and the cloud-free image is of great significance for land use classification, change detection, and long time-series landcover monitoring. Traditional cloud detection methods based on spectral and texture features have acquired certain effects in complex scenarios, such as cloud–snow mixing, but there is still a large room for improvement in terms of generation ability. In recent years, cloud detection with deep-learning methods has significantly improved the accuracy in complex regions such as high-brightness feature mixing areas. However, the existing deep learning-based cloud detection methods still have certain limitations. For instance, a few omission alarms and commission alarms still exist in cloud edge regions. At present, the cloud detection methods based on deep learning are gradually converted from a pure convolutional structure to a global feature extraction perspective, such as attention modules, but the computational burden is also increased, which is difficult to meet for the rapidly developing time-sensitive tasks, such as onboard real-time cloud detection in optical remote sensing imagery. To address the above problems, this manuscript proposes a high-precision cloud detection network fusing a self-attention module and spatial pyramidal pooling. Firstly, we use the DenseNet network as the backbone, then the deep semantic features are extracted by combining a global self-attention module and spatial pyramid pooling module. Secondly, to solve the problem of unbalanced training samples, we design a weighted cross-entropy loss function to optimize it. Finally, cloud detection accuracy is assessed. With the quantitative comparison experiments on different images, such as Landsat8, Landsat9, GF-2, and Beijing-2, the results indicate that, compared with the feature-based methods, the deep learning network can effectively distinguish in the cloud–snow confusion-prone region using only visible three-channel images, which significantly reduces the number of required image bands. Compared with other deep learning methods, the accuracy at the edge of the cloud region is higher and the overall computational efficiency is relatively optimal.

The systematic monitoring of the Earth using optical satellites is limited by the presence of clouds. Accurately detecting these clouds is necessary to exploit satellite image archives in remote sensing applications. Despite many developments, cloud detection remains an unsolved problem with room for improvement, especially over bright surfaces and thin clouds. Recently, advances in cloud masking using deep learning have shown significant boosts in cloud detection accuracy. However, these works are validated in heterogeneous manners, and the comparison with operational threshold-based schemes is not consistent among many of them. In this work, we systematically compare deep learning models trained on Landsat-8 images on different Landsat-8 and Sentinel-2 publicly available datasets. Overall, we show that deep learning models exhibit a high detection accuracy when trained and tested on independent images from the same Landsat-8 dataset (intra-dataset validation), outperforming operational algorithms. However, the performance of deep learning models is similar to operational threshold-based ones when they are tested on different datasets of Landsat-8 images (inter-dataset validation) or datasets from a different sensor with similar radiometric characteristics such as Sentinel-2 (cross-sensor validation). The results suggest that (i) the development of cloud detection methods for new satellites can be based on deep learning models trained on data from similar sensors and (ii) there is a strong dependence of deep learning models on the dataset used for training and testing, which highlights the necessity of standardized datasets and procedures for benchmarking cloud detection models in the future.

Accurate cloud detection is critical for advancing atmospheric monitoring and meteorological forecasting. This paper presents the Cloud Detection Challenge, an initiative aimed at enhancing cloud detection through innovative solutions using lidar-based ceilometer data. This initiative was hosted by IEEE MetroXRAINE 2024, and 11 teams participated in this initiative. Participants were provided with a novel dataset of backscatter profiles converted into time-height plots, offering unique insights into atmospheric conditions beyond conventional imagery. Data collection employed a Lufft CHM 15k ceilometer, capturing cloud dynamics every 15 seconds located near Mt. Etna, an active volcano in Italy. The dataset includes 1568 hourly labeled backscatter profiles, serving as a benchmark for state-of-the-art deep learning models. The challenge sets a baseline performance of 89.57% accuracy, 92.73% F1-score, 89.82% precision, and 95.84% recall, inviting participants to develop models to exceed these results. Submissions proposed a wide-range of AI-based approaches, including Transformer and Convolutional Neural Network architectures, showcasing the potential of advanced image analysis techniques in lidar-based cloud detection. This paper details the challenge framework, as well as the methodologies proposed by top-performing teams, offering a comparative evaluation of their effectiveness. Our initiative advances cloud detection technologies and underscores their broader implications for environmental monitoring, agriculture, and satellite imaging. The insights and dataset presented herein lay the groundwork for future advancements in leveraging lidar data for atmospheric analysis.

The existence of clouds has seriously affected the application of remote sensing data. Therefore, accurate cloud detection is of great significance in remote sensing image processing and application. Traditional cloud detection methods are complex to operate and often require the additional ancillary information. An automatic cloud detection method based on convolutional neural network (CNN) is proposed in this study. The method utilizes a convolutional network structure to classify training samples for cloud and non-cloud. In order to make full use of image information, images of different band numbers are applied to evaluate the influence of the spectrum on cloud detection. Experiments and verification on Landsat 8 images show that the proposed method based on CNN can comprehensively and automatically detect different types of clouds on different surface types, and the cloud detection result using 7 bands is the optimal. The algorithm takes full advantage of image information and does not rely on thermal infrared information, which has practical application value for improving image utilization and subsequent retrieval of remote sensing parameters.

Remote sensing is an indispensable technical way for monitoring earth resources and environmental changes. However, optical remote sensing images often contain a large number of cloud, especially in tropical rain forest areas, make it difficult to obtain completely cloud-free remote sensing images. Therefore, accurate cloud detection is of great research value for optical remote sensing applications. In this paper, we propose a saliency model-oriented convolution neural network for cloud detection in remote sensing images. Firstly, we adopt Kernel Principal Component Analysis (KCPA) to unsupervised pre-training the network. Secondly, small labeled samples are used to fine-tune the network structure. And, remote sensing images are performed with super-pixel approach before cloud detection to eliminate the irrelevant backgrounds and non-clouds object. Thirdly, the image blocks are input into the trained convolutional neural network (CNN) for cloud detection. Meanwhile, the segmented image will be recovered. Fourth, we fuse the detected result with the saliency map of raw image to further improve the accuracy of detection result. Experiments show that the proposed method can accurately detect cloud. Compared to other state-of-the-art cloud detection method, the new method has better robustness.

Accurate cloud detection using medium-resolution multispectral satellite imagery (such as Landsat and Sentinel data) is always difficult due to the complex land surfaces, diverse cloud types, and limited number of available spectral bands, especially in the case of images without thermal bands. In this paper, a novel classification extension-based cloud detection (CECD) method was proposed for masking clouds in the medium-resolution images. The new method does not rely on thermal bands and can be used for masking clouds in different types of medium-resolution satellite imagery. First, with the support of low-resolution satellite imagery with short revisit periods, cloud and non-cloud pixels were identified in the resampled low-resolution version of the medium-resolution cloudy image. Then, based on the identified cloud and non-cloud pixels and the resampled cloudy image, training samples were automatically collected to develop a random forest (RF) classifier. Finally, the developed RF classifier was extended to the corresponding medium-resolution cloudy image to generate an accurate cloud mask. The CECD method was applied to Landsat-8 and Sentinel-2 imagery to test the performance for different satellite images, and the well-known function of mask (FMASK) method was employed for comparison with our method. The results indicate that CECD is more accurate at detecting clouds in Landsat-8 and Sentinel-2 imagery, giving an average F-measure value of 97.65% and 97.11% for Landsat-8 and Sentinel-2 imagery, respectively, as against corresponding results of 90.80% and 88.47% for FMASK. It is concluded, therefore, that the proposed CECD algorithm is an effective cloud-classification algorithm that can be applied to the medium-resolution optical satellite imagery.