Cloud Detection Research Articles

A Sea Ice Recognition Algorithm in Bohai Based on Random Forest

As an important maritime hub, Bohai Sea Bay provides great convenience for shipping and suffers from sea ice disasters of different severity every winter, which greatly affects the socio-economic and development of the region. Therefore, this paper uses FY-4A (a weather satellite) data to study sea ice in the Bohai Sea. After processing the data for land removal and cloud detection, it combines multi-channel threshold method and adaptive threshold algorithm to realize the recognition of Bohai Sea ice under clear sky conditions. The random forests classification algorithm is introduced in sea ice identification, which can achieve a certain effect of sea ice classification recognition under cloud cover. Under non-clear sky conditions, the results of Bohai Sea ice identification based on random forests have been improved, and the algorithm can effectively identify Bohai Sea Ice and can improve the accuracy of sea ice identification, which lays a foundation for the accuracy and stability of sea ice identification. It realizes sea ice identification in the Bohai Sea and provides data support and algorithm support for marine climate forecasting related departments.

3D Cascade RCNN: High Quality Object Detection in Point Clouds.

Recent progress on 2D object detection has featured Cascade RCNN, which capitalizes on a sequence of cascade detectors to progressively improve proposal quality, towards high-quality object detection. However, there has not been evidence in support of building such cascade structures for 3D object detection, a challenging detection scenario with highly sparse LiDAR point clouds. In this work, we present a simple yet effective cascade architecture, named 3D Cascade RCNN, that allocates multiple detectors based on the voxelized point clouds in a cascade paradigm, pursuing higher quality 3D object detector progressively. Furthermore, we quantitatively define the sparsity level of the points within 3D bounding box of each object as the point completeness score, which is exploited as the task weight for each proposal to guide the learning of each stage detector. The spirit behind is to assign higher weights for high-quality proposals with relatively complete point distribution, while down-weight the proposals with extremely sparse points that often incur noise during training. This design of completeness-aware re-weighting elegantly upgrades the cascade paradigm to be better applicable for the sparse input data, without increasing any FLOP budgets. Through extensive experiments on both the KITTI dataset and Waymo Open Dataset, we validate the superiority of our proposed 3D Cascade RCNN, when comparing to state-of-the-art 3D object detection techniques. The source code is publicly available at https://github.com/caiqi/Cascasde-3D.

Implementation of Data Mining on a Secure Cloud Computing over a Web API using Supervised Machine Learning Algorithm

Ever since the era of internet had ushered in cloud computing, there had been increase in the demand for the unlimited data available through cloud computing for data analysis, pattern recognition and technology advancement. With this also bring the problem of scalability, efficiency and security threat. This research paper focuses on how data can be dynamically mine in real time for pattern detection in a secure cloud computing environment using combination of decision tree algorithm and Random Forest over a restful Application Programming Interface (API). We are able to successfully Implement data mining on cloud computing bypassing or avoiding direct interaction with data warehouse and without any terminal involve by using combination of IBM Cloud storage facility, Amazing Web Service, Application Programming Interface and Window service along with a decision tree and Random Forest algorithm for our classifier. We were able to successfully bypass direct connection with the data warehouse and cloud terminal with 94% accuracy in our model.

Effect of the Instrument Slit Function on Upwelling Radiance from a Wavelength Dependent Surface Reflectance

The Radiance Enhancement (RE) method was introduced for efficient detection of clouds from the space. Recently, we have also reported that due to high reflectance of combustion-originated smokes, this approach can also be generalized for detection of the forest fires by retrieving and analyzing datasets collected from a space orbiting micro-spectrometer operating in the near infrared spectral range. In our previous publication, we have performed a comparison of observed and synthetic radiance spectra by developing a method for computation of surface reflectance consisting of different canopies by weighted sum based on their areal coverage. However, this approach should be justified by a method based on corresponding proportions of the upwelling radiance. The results of computations we performed in this study reveal a good match between areal coverage of canopies and the corresponding proportions of the upwelling radiance due to effect of the instrument slit function.

Retrieval of Sea Surface Temperature From HY-1B COCTS

The Chinese Ocean Color and Temperature Scanner (COCTS) on board HY-1 series satellites has two thermal infrared channels with the spectrum range of 10.30-11.40 μm and 11.40-12.50 μm for sea surface temperature (SST) observations. To reprocess the Haiyang-1B (HY-1B) COCTS SST, the Bayesian cloud detection and optimal estimation (OE) SST retrieval were applied to COCTS data in this study. The Bayesian cloud detection algorithm that has been developed is based on the Bayes’ theorem and uses simulation of COCTS observations. The MODerate resolution atmospheric TRANsmission (MODTRAN) model was used for simulation of COCTS brightness temperatures. SSTs were retrieved from COCTS by OE from 2009 to 2011 in the northwest Pacific. Comparison of COCTS OE SST with in situ SST showed that the COCTS SSTs are cooler than buoy measurements by –0.23 °C on average, and the standard deviation (SD) of differences was 0.51 °C. A large component of the mean difference is attributable to the cool skin effect at the ocean surface (typically –0.15 to –0.2 °C), the remainder being attributable to simulation and calibration biases. The mean difference of COCTS OE SST with matched skin temperatures from the Advanced Along Track Scanning Radiometer (AATSR) is closer to zero, being –0.09 °C, with a SD of 0.49 °C. These validation results of COCTS OE SST demonstrate that Bayesian cloud detection and OE SST retrieval algorithm work well for improving COCTS SST accuracy, and show the potential of these methods to help develop SST products for operational HY-1 satellites, HY-1C and HY-1D.

Temporal Shape-Based Fusion Method to Generate Continuous Vegetation Index at Fine Spatial Resolution

In this study, a temporal shape–based fusion method using a spatially and temporally moving window is proposed to incorporate time lag of fine and coarse resolution observations, and to fully utilize target fine resolution pixel and similar coarse resolution pixels in the process. This method provides high accuracy fused images with Pearson’s r of ~0.95, root mean square error of ~0.04, and bias of ~0.01 for commonly used fine spatial resolution satellites, including Landsat 7 and 8, Sentinel 2, and Gaofen 1, over different heterogeneous regions, such as urban, mountain, forest, and savanna regions. The fused fine resolution Enhanced Vegetation Index (EVI) time series using different fine spatial resolution satellites data as input are all highly correlated with the PhenoCam monitored green chromatic coordinate, with no temporal lag. Compared with commonly used data fusion method, this method provides equivalent and slightly higher accuracy because both neighboring similar pixels and the annual temporal variation are fully considered. This temporal shape–based fusion method does not require each input fine resolution image to be cloud-free; therefore, it can be used at a large spatial scale without further preprocessing and generates continuous datasets over a long-time range with only one input preparation process. The factors that could affect the method accuracy are the cloud detection accuracy of fine resolution data and the temporal continuity of the coarse resolution data. The method may also be used to produce spatially and temporally continuous surface reflectance and other surface reflectance derived indices.

Cloud Detection Algorithm of Micro-Pulse Lidar Based on Bidirectional Reconstruction of Backscatter Signal

Cloud Detection Algorithm of Micro-Pulse Lidar Based on Bidirectional Reconstruction of Backscatter Signal

The Φ-Sat-1 Mission: The First On-Board Deep Neural Network Demonstrator for Satellite Earth Observation

Artificial intelligence (AI) is paving the way for a new era of algorithms focusing directly on the information contained in the data, autonomously extracting relevant features for a given application. While the initial paradigm was to have these applications run by a server hosted processor, recent advances in microelectronics provide hardware accelerators with an efficient ratio between computation and energy consumption, enabling the implementation of AI algorithms &#x201C;at the edge.&#x201D; In this way only the meaningful and useful data are transmitted to the end-user, minimizing the required data bandwidth, and reducing the latency with respect to the cloud computing model. In recent years, European Space Agency (ESA) is promoting the development of disruptive innovative technologies on-board earth observation (EO) missions. In this field, the most advanced experiment to date is the <inline-formula> <tex-math notation="LaTeX">$\Phi $ </tex-math></inline-formula>-sat-1, which has demonstrated the potential of artificial intelligence (AI) as a reliable and accurate tool for cloud detection on-board a hyperspectral imaging mission. The activities involved included demonstrating the robustness of the Intel Movidius Myriad 2 hardware accelerator against ionizing radiation, developing a Cloudscout segmentation neural network (NN), run on Myriad 2, to identify, classify, and eventually discard on-board the cloudy images, and assessing the innovative Hyperscout-2 hyperspectral sensor. This mission represents the first official attempt to successfully run an AI deep convolutional NN (CNN) directly inferencing on a dedicated accelerator on-board a satellite, opening the way for a new era of discovery and commercial applications driven by the deployment of on-board AI.

Development of a new hybrid cloud cover detection method for all sky mapping

Local precise cloud detection techniques in the daytime are finding a wide range of applications in solar farms and for tracking the factory flue gases. Additionally, the cloud coverage is a measure for the site quality of astronomical observatories. The moon is itself a problem for astronomical observations. If there are astronomical objects near to moon, the light scattered by clouds from moon in the sky also has to be determined to extract from the observation schedule of robotic telescope. A novel hybrid method to deal with all-sky images in any conditions in daytime and nighttime is developed, and related algorithms are implemented in MATLAB. A GUI has been presented which interactively analyzes the images taken from all-sky cams. It plots a radar scan of the sky with clouds in polar coordinate systems. The systematic mean error is estimated at less than 3%.

Robust Thick Cloud Removal for Multitemporal Remote Sensing Images Using Coupled Tensor Factorization

The existing nonblind cloud and cloud shadow (cloud/shadow) removal methods for remote sensing (RS) images are based on the assumption that cloud/shadow masks are accurately given. Since the masks are usually manually labeled or detected by cloud detection methods, whose accuracy cannot be well guaranteed, the cloud/shadow removal effect may be affected. In this article, we suggest a robust thick cloud/shadow removal (RTCR) method that meets the problem with an inaccurate mask. To faithfully reconstruct the multitemporal information, a coupled tensor factorization is used to explore the relationship between the abundances of the multitemporal images in the same scene. Moreover, an efficient algorithm is developed to solve the proposed model based on the augmented Lagrange multiplier method. The experimental results under accurate masks and inaccurate masks demonstrate its robustness and superiority for thick cloud/shadow removal.

TransCloudSeg: Ground-Based Cloud Image Segmentation With Transformer

Cloud image segmentation plays an important role in ground-based cloud observation. Recently, most existing methods for ground-based cloud image segmentation learn feature representations using the convolutional neural network (CNN), which results in the loss of global information because of the limited receptive field size of the filters in the CNN. In this article, we propose a novel deep model named TransCloudSeg, which makes full use of the advantages of the CNN and transformer to extract detailed information and global contextual information for ground-based cloud image segmentation. Specifically, TransCloudSeg hybridizes the CNN and transformer as the encoders to obtain different features. To recover and fuse the feature maps from the encoders, we design the CNN decoder and the transformer decoder for TransCloudSeg. After obtaining two sets of feature maps from two different decoders, we propose the heterogeneous fusion module to effectively fuse the heterogeneous feature maps by applying the self-attention mechanism. We conduct a series of experiments on Tianjin Normal University large-scale cloud detection database and Tianjin Normal University cloud detection database, and the results show that our method achieves a better performance than other state-of-the-art methods, thus proving the effectiveness of the proposed TransCloudSeg.

Category Correlation and Adaptive Knowledge Distillation for Compact Cloud Detection in Remote Sensing Images

Category Correlation and Adaptive Knowledge Distillation for Compact Cloud Detection in Remote Sensing Images

Cloud detection for FY-3D MERSI &lt;bold&gt;Ⅱ&lt;/bold&gt; images combine radiative transfer simulation and shallow neural network

Cloud detection for FY-3D MERSI &lt;bold&gt;Ⅱ&lt;/bold&gt; images combine radiative transfer simulation and shallow neural network

Detecting Multilayer Clouds From the Geostationary Advanced Himawari Imager Using Machine Learning Techniques

This study develops a machine learning (ML)-based multilayer cloud detection algorithm for the passive Advanced Himawari Imager (AHI) aboard the geostationary Himawari-8 satellite. AHI measurements in the 0.64-, 1.6-, 2.3-, 3.9-, 7.3-, 8.6-, 11.2-, and 12.4- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula> channels, their combinations, geolocations, and observational geometries are used as predictors, and collocated active CloudSat and CALIPSO data are used to accurately label multilayer cloud pixels as the reference/truth of the predictand. We develop an ML-based daytime model (ML-Day) that utilizes all the aforementioned predictors and an all-time one (ML-All) that excludes the solar channel-dependent variables. Among four ML algorithms, the random forest (RF) performs slightly better than the artificial neural network, K-nearest neighbor, and support vector machines. By comparing with the merged CloudSat and CALIPSO product, the ML-Day model correctly identifies ~89% single-layer clouds and ~70% multilayer clouds, outperforming the Moderate Resolution Imaging Spectroradiometer (MODIS) operational multilayer cloud product (~80% and ~40% given by Marchant <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">et al.</i> ). The success rates of ML-All for single-layer and multilayer clouds also reach ~85% and ~64%, respectively. The misclassification of our algorithm is mostly caused by missing optically thin clouds, a drawback of most radiometers without the 1.38- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula> channel. Furthermore, with multilayer cloud pixels well detected by our algorithm, the AHI operational cloud top height retrievals are found to be larger biased due to multilayer cloud occurrence and might be improved by considering cloud vertical structures.

Estimating 250-m Land Surface and Atmospheric Variables From MERSI Top-of-Atmosphere Reflectance

The medium-resolution spectral imager (MERSI) onboard China’s new generation of polar-orbiting meteorological satellite series FengYun-3 (FY-3) is providing global observations of the Earth’s environment. One of the important characteristics of the MERSI sensor is that it has four bands at 250-m spatial resolution. However, very few high-level land surface and atmospheric products have been generated from MERSI data. This article presents an effective optimization method to estimate the continuous temporal distribution of multiple land surface and atmospheric variables at a 250-m spatial resolution from time series FY-3B MERSI top-of-atmosphere (TOA) observations. The MODIS cloud detection method was applied to the MERSI TOA reflectance to determine the clear observations. Then, a physics-based BRDF correction was adopt to reduce the topographic effects using the slope angle and aspect angle of the slope. Finally, a coupled land surface-atmosphere radiative transfer (RT) model and the shuffled complex evolution (SCE) optimization algorithm were used to estimate a suit of variables, including the leaf area index (LAI), the aerosol optical depth (AOD), the cloud optical thickness (COT), the cloud effective particle radius (CER), the land surface reflectance, the shortwave and visible albedo, the incident shortwave radiation (ISR), the incident photosynthetically active radiation (PAR), the fraction of absorbed PAR (FAPAR), the surface broadband emissivity (BBE), and the TOA shortwave albedo. The method was applied to the data from the 18 SURFRAD, the America FluxNet (AmeriFlux), and the Images and Beijing Normal University net (BNUnet) sites. All the estimated variables were also compared with Moderate Resolution Imaging Spectroradiometer (MODIS) and Global Land Surface Satellite (GLASS) products. The estimated LAI, PAR, FAPAR, shortwave albedo, and ISR were also validated using ground measurements from the selected sites. The results show that our method can simultaneously estimate multiple temporally continuous variables in both clear-and cloudy-sky conditions, with an accuracy comparable to those of MODIS and GLASS products. The estimated LAI, PAR, FAPAR, shortwave albedo, and ISR are consistent with the field measurements, with coefficients of determination ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> ) of 0.692, 0.783, 0.788, 0.559, and 0.833, and root mean square errors (RMSEs) of 0.427, 56.681 W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , 0.092, 0.084, and 115.305 W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> , respectively. The proposed algorithm can effectively estimate the instantaneous atmospheric and land surface variables directly from satellite observations without cumbersome atmospheric corrections. Moreover, it can potentially be applied to multiple satellite observations.

Gated Aggregation Network for Cloud Detection in Remote Sensing Image

Gated Aggregation Network for Cloud Detection in Remote Sensing Image

Improving Deep Learning-Based Cloud Detection for Satellite Images With Attention Mechanism

Improving Deep Learning-Based Cloud Detection for Satellite Images With Attention Mechanism

CNN Cloud Detection Algorithm Based on Channel and Spatial Attention and Probabilistic Upsampling for Remote Sensing Image

In the field of remote sensing image, how to transmit image information more efficiently with limited bandwidth has always been a research hotspot. Compared with other ground objects, cloud pixels in remote sensing image are invalid information, so it is a meaningful research work to remove cloud before transmitting image and reduce the waste of useless information. In remote sensing image, due to the existence of thin clouds and the complexity of the underlying surface, most of the cloud detection algorithms struggle to achieve effective separation of clouds and ground objects. A deep learning (DL) cloud detection algorithm based on attention mechanism and probability upsampling has been proposed in this article. In order to enhance the information of the key areas, in the channel attention module, crucial information is highlighted in the channel dimension of the encoder, and the useless information is weakened. The spatial attention module is in the spatial dimension. The information fusion between each point in the image is strengthened. To reduce the information loss caused by the down-sampling module, a probabilistic upsampling block (PUB) is proposed to restore the image. Eventually, experiments are performed on Gaofen-1WFV data, and the results indicate that the algorithm proposed in this article has better detection results than other cloud detection algorithms in different scenarios.

Multifractal analysis for spatial characterization of high resolution Sentinel-2/MAJA products in Southwestern France

Multifractal analysis was applied to evidence spatial properties, in particular scaling behaviour, of Sentinel-2 surface reflectances and optical indexes acquired over the Southwest region of France. Considering the significant impact of clouds on the spatial properties of optical products, no study to our knowledge has focused on the multifractal analysis of a series of images. Images with minimal cloud cover were identified for the whole year 2016 based on the MACCS-ATCOR Joint Algorithm (MAJA) cloud and shadow detection process. This allowed us to access the dynamics of the scaling properties of surface reflectances, contributing to demonstrate the potential of multifractals for spatial characterization. Different scaling regimes, separated by significant scale breaks, were evident in these data. Through simulations from the Universal Multifractal model, we could explain the scaling properties and especially the position of scale breaks observed in the images studied by two types of phenomena. First, a transition scale located between hundreds of meters and one kilometre was detected on power spectra, structure functions and statistical moments. This specific scale was related to the average size of the agricultural plots of the studied region, revealing the direct link between the scaling behaviour observed and the physical properties of the surface. Second, an instrumental phenomenon was detected on the power spectra, resulting in a systematic transition scale observed near the resolution of the sensor and revealing the impact of the acquisition system on the high frequencies of optical images. In this research work, theoretical and/or empirical validation arguments were provided as to the origin of the scales detected, contributing to demonstrate the value of multifractals to extract image texture from surface reflectance products.

PANDA: parallel asymmetric network with double attention for cloud and its shadow detection

In remote sensing images, the complex and changeable background often interferes with the detection of cloud and its shadow, which leads to the phenomenon of missing detection in areas with similar background colors. In addition, most methods cannot achieve the real-time detection effect because of too much calculation. To solve the above problems, the parallel asymmetric network with double attention is proposed. The algorithm adopts a parallel method, which allows two branches to participate in the calculation at the same time. One branch is used to fuse different information from two branches at different levels, and another branch is responsible for extracting deeper context information. At the same time, double attention module and asymmetric dilated block are used for two branches, respectively. Double attention module can help the algorithm pay more attention to the category and spatial information of cloud and its shadow, thus reducing the interference caused by background information in images. Asymmetric dilated block can extract two levels of receptive fields, and it can help the network to obtain enough receptive fields in cloud and its shadow images, thus reducing the cases of missed detection and false detection. Moreover, these two modules are lightweight in their parameters and calculation. Compared with some previous methods, our method can guarantee the accuracy and the detection speed is fast.