Cloud Shadows Research Articles

KappaMask: AI-Based Cloudmask Processor for Sentinel-2

The Copernicus Sentinel-2 mission operated by the European Space Agency (ESA) provides comprehensive and continuous multi-spectral observations of all the Earth’s land surface since mid-2015. Clouds and cloud shadows significantly decrease the usability of optical satellite data, especially in agricultural applications; therefore, an accurate and reliable cloud mask is mandatory for effective EO optical data exploitation. During the last few years, image segmentation techniques have developed rapidly with the exploitation of neural network capabilities. With this perspective, the KappaMask processor using U-Net architecture was developed with the ability to generate a classification mask over northern latitudes into the following classes: clear, cloud shadow, semi-transparent cloud (thin clouds), cloud and invalid. For training, a Sentinel-2 dataset covering the Northern European terrestrial area was labelled. KappaMask provides a 10 m classification mask for Sentinel-2 Level-2A (L2A) and Level-1C (L1C) products. The total dice coefficient on the test dataset, which was not seen by the model at any stage, was 80% for KappaMask L2A and 76% for KappaMask L1C for clear, cloud shadow, semi-transparent and cloud classes. A comparison with rule-based cloud mask methods was then performed on the same test dataset, where Sen2Cor reached 59% dice coefficient for clear, cloud shadow, semi-transparent and cloud classes, Fmask reached 61% for clear, cloud shadow and cloud classes and Maja reached 51% for clear and cloud classes. The closest machine learning open-source cloud classification mask, S2cloudless, had a 63% dice coefficient providing only cloud and clear classes, while KappaMask L2A, with a more complex classification schema, outperformed S2cloudless by 17%.

Real-Time Automated Classification of Sky Conditions Using Deep Learning and Edge Computing

The radiometric quality of remotely sensed imagery is crucial for precision agriculture applications because estimations of plant health rely on the underlying quality. Sky conditions, and specifically shadowing from clouds, are critical determinants in the quality of images that can be obtained from low-altitude sensing platforms. In this work, we first compare common deep learning approaches to classify sky conditions with regard to cloud shadows in agricultural fields using a visible spectrum camera. We then develop an artificial-intelligence-based edge computing system to fully automate the classification process. Training data consisting of 100 oblique angle images of the sky were provided to a convolutional neural network and two deep residual neural networks (ResNet18 and ResNet34) to facilitate learning two classes, namely (1) good image quality expected, and (2) degraded image quality expected. The expectation of quality stemmed from the sky condition (i.e., density, coverage, and thickness of clouds) present at the time of the image capture. These networks were tested using a set of 13,000 images. Our results demonstrated that ResNet18 and ResNet34 classifiers produced better classification accuracy when compared to a convolutional neural network classifier. The best overall accuracy was obtained by ResNet34, which was 92% accurate, with a Kappa statistic of 0.77. These results demonstrate a low-cost solution to quality control for future autonomous farming systems that will operate without human intervention and supervision.

Towards a novel approach for Sentinel-3 synergistic OLCI/SLSTR cloud and cloud shadow detection based on stereo cloud-top height estimation

Sentinel-3 is an Earth observation satellite constellation launched by the European Space Agency. Each satellite carries two optical multispectral instruments: the Ocean and Land Colour Instrument (OLCI) and the Sea and Land Surface Temperature Radiometer (SLSTR). OLCI and SLSTR sensors produce images covering the visible and infrared spectrum that can be collocated in order to generate synergistic products. In Earth observation, a particular weakness of optical sensors is their high sensitivity to clouds and their shadows. An incorrect cloud and cloud shadow detection leads to mistakes in both land and ocean retrievals of biophysical parameters. In order to exploit both OLCI and SLSTR capabilities, image co-registration at ground level is needed. However, applying such collocation of the images results in cloud location mismatches due to the different viewing angles of OLCI and SLSTR, which complicates the synergistic cloud detection. This study seeks to provide a solution to correctly obtain the projected clouds based on the estimation of cloud top heights in order to better collocate clouds between sensors and detect their shadows. The study presents a forward and backward method to estimate the real nadir position of a cloud on the satellite image starting from an existing cloud mask, as well as the corresponding cloud projections on the surface depending on the solar and sensor viewing angles. The estimation of cloud top heights is based on differences in the cloud projections from SLSTR nadir and oblique views. Experimental results show that the stereo cloud matching based on maximum cross-correlation between SLSTR nadir and oblique spectra was the most robust method to match SLSTR clouds for both nadir and oblique views as compared to spectral distance and spectral angle minimization. We test the method over several images around the world, leading to higher overall accuracy (OA) as compared to Sentinel-3 official products, both in detecting SLSTR clouds and OLCI cloud shadows (SLSTR nadir OA = 93.6%, SLSTR oblique OA = 88.7%, OLCI cloud shadow OA = 93.9% for the stereo matcher, against 82.2%, 81.3% and 90.5%, respectively, for the official Sentinel-3 products). This study also provides a starting point in the development of a cloud screening approach for the upcoming Fluorescence Explorer (FLEX) satellite mission, expected to fly in tandem with Sentinel-3.

Prediction of Solar Power Using Near-Real Time Satellite Data

Solar energy production is affected by the attenuation of incoming irradiance from underlying clouds. Often, improvements in the short-term predictability of irradiance using satellite irradiance models can assist grid operators in managing intermittent solar-generated electricity. In this paper, we develop and test a satellite irradiance model with short-term prediction capabilities using cloud motion vectors. Near-real time visible images from Himawari-8 satellite are used to derive cloud motion vectors using optical flow estimation techniques. The cloud motion vectors are used for the advection of pixels at future time horizons for predictions of irradiance at the surface. Firstly, the pixels are converted to cloud index using the historical satellite data accounting for clear, cloudy and cloud shadow pixels. Secondly, the cloud index is mapped to the clear sky index using a historical fitting function from the respective sites. Thirdly, the predicated all-sky irradiance is derived by scaling the clear sky irradiance with a clear sky index. Finally, a power conversion model trained at each site converts irradiance to power. The prediction of solar power tested at four sites in Australia using a one-month benchmark period with 5 min ahead prediction showed that errors were less than 10% at almost 34–60% of predicted times, decreasing to 18–26% of times under live predictions, but it outperformed persistence by >50% of the days with errors <10% for all sites. Results show that increased latency in satellite images and errors resulting from the conversion of cloud index to irradiance and power can significantly affect the forecasts.

The potential of active and passive remote sensing to detect frequent harvesting of alfalfa

Alfalfa (Medicago sativa L.), referred to as the “Queen of Forages” because of its importance among forage crops, provides high quality forage for the livestock industry. The timing and frequency of alfalfa hay harvesting have implications on its quality and quantity. With ever-increasing capability, it is possible to use satellite remote sensing data to monitor alfalfa harvests. This study investigated the potential of using satellite remote sensing to capture frequent harvesting events on an alfalfa field in central Oklahoma. Both passive remote sensing data, namely Moderate Resolution Imaging Spectroradiometer (MODIS), Landsat-7 and -8, Sentinel-2, Harmonized Landsat and Sentinel-2 (HLS), and active remote sensing data, namely Sentinel-1, were included. Our results indicate that good quality optical remote sensing datasets (i.e., cloud and cloud shadow free) with both fine spatial (≤100 m) and high temporal (effective observation at 8-day intervals or better) resolutions are necessary to detect frequent alfalfa harvesting events, challenged by possible adverse weather conditions and quick regrowth of vegetation after harvest. Landsat (7 and 8) and Sentinel-2 were more sensitive to changes in vegetation indices after harvest than MODIS due to their higher spatial resolutions, which helped avoid the mixed pixel issue in MODIS caused by its coarser spatial resolution (∼500 m). Combining Landsat (7 and 8) with Sentinel-2 imageries through linear regression between the Normalized Difference Vegetation Index (NDVI) values, up to one week apart, increased the accuracy of detecting frequent alfalfa harvesting events. The responses of HLS to alfalfa harvesting events were similar with fused Landsat and Sentinel-2 data using their linear relationship of NDVI values. However, the high noise level in the HLS data needs to be minimized before it can be used to detect alfalfa harvests at the regional scale. In most cases, both Sentinel-1 radar backscatter coefficients (vertical transmit and vertical receive, VV + vertical transmit and horizontal receive, VH) and interferometric coherence from Sentinel-1 Simple Look Complex (SLC) data were decreased by harvesting events in small incident angle observations (34.31°). No consistent relationships existed between backscatter or coherence and alfalfa harvests in larger incident angle observations (45.11°). Future studies should focus on small incident angle observations instead of processing all of the radar data, which has big data volume and is time-consuming. Overall, active radar has the potential to detect alfalfa harvesting events. However, it is visually less intuitive than optical data with incident angles, quantity harvested, and soil moisture being the compounding factors. This study illustrates that combining multiple optical sensors with a fine spatial resolution (e.g., Landsat-7, 8, and Sentinel-2) and/or fusing radar with optical remote sensing to increase the temporal resolution are promising approaches to detect frequent alfalfa harvesting events and other hay harvesting activities.

Strip pooling channel spatial attention network for the segmentation of cloud and cloud shadow

The background in image of remote sensing is often complicated and changeable, and the edge of cloud and its shadow is irregular. In the traditional method, the bright part of the background is easy to be misjudged as cloud, while the dark part is easy to be misjudged as cloud shadow. Moreover, the edge information of the extracted cloud and its shadow is rough, and it is easy to miss the judgment for the thin cloud part and the light cloud shadow part. In order to solve the above problems, a strip pooling channel spatial attention network is proposed. In this work, the strip pooling residual network is used as the backbone network to obtain the feature of cloud and its shadow. The strip pooling residual network can obtain more accurate local position information of cloud and its shadow, which can improve the accuracy of edge segmentation. Channel attention and spatial attention combine shallow spatial information with deep context information, so that cloud and its shadow can be accurately segmented from the background. The experimental results demonstrate that method in our work can acquire more accurate segmentation edge than existing methods, hence it is practical in accurate cloud and its shadow segmentation.

Guaranteed Robust Tensor Completion via ∗L-SVD with Applications to Remote Sensing Data

This paper conducts a rigorous analysis for the problem of robust tensor completion, which aims at recovering an unknown three-way tensor from incomplete observations corrupted by gross sparse outliers and small dense noises simultaneously due to various reasons such as sensor dead pixels, communication loss, electromagnetic interferences, cloud shadows, etc. To estimate the underlying tensor, a new penalized least squares estimator is first formulated by exploiting the low rankness of the signal tensor within the framework of tensor ∗L-Singular Value Decomposition (∗L-SVD) and leveraging the sparse structure of the outlier tensor. Then, an algorithm based on the Alternating Direction Method of Multipliers (ADMM) is designed to compute the estimator in an efficient way. Statistically, the non-asymptotic upper bound on the estimation error is established and further proved to be optimal (up to a log factor) in a minimax sense. Simulation studies on synthetic data demonstrate that the proposed error bound can predict the scaling behavior of the estimation error with problem parameters (i.e., tubal rank of the underlying tensor, sparsity of the outliers, and the number of uncorrupted observations). Both the effectiveness and efficiency of the proposed algorithm are evaluated through experiments for robust completion on seven different types of remote sensing data.

VALIDATION OF SOIL USES AROUND RESERVOIRS IN THE SEMI-ARID THROUGH IMAGE CLASSIFICATION

ABSTRACT The work evaluated the potential for discrimination of land use and occupation around reservoirs, using spectral information obtained by multispectral, hyperspectral satellites and images obtained with an ARP (remotely piloted aircraft). The research analyzed the performance of different images classification techniques applied to multispectral and hyperspectral sensors for the detection and differentiation of soil classes around the Paus Brancos and Marengo reservoirs, located in Settlement 25 of Maio. The classes identified based on surveys in campaigns carried out in 2014 and 2015 around the reservoirs were: water, macrophytes, exposed soil, native vegetation, agriculture, thin and ebbing vegetation, in addition to the cloud and cloud shadow targets. The performance of the classifiers applied to the image of the Hyperion sensor was, in general, superior to those obtained in Landsat 8 image, which can be explained by the high spectral resolution of the first, which facilitates the differentiation of targets with similar spectral response. For validation of the supervised classification method of Maximum Likelihood, Landsat 8 (08/24/2015) and Hyperion (08/28/2015) images were used. The results of the application indicated a good performance of the classifier associated with the RGB composition of the chosen Hyperion image (bands R - 51, G - 161, B - 19) in the detection of the classes around this reservoir, producing a Kappa coefficient of 0.83. The availability of data from the Hyperion sensor is very restricted, which hinders the development of continued research, thus the use of images surpassed by RPA is extremely viable.

Global clear sky near-surface imagery from multiple satellite daily imagery time series

We develop a new statistically-robust adaptive regression method (SARM) to extract clear sky true color imagery, approximating the near-surface imagery, derived from multiple satellite daily imagery time series, while avoiding artifacts due to clouds and cloud shadows. We compare the SARM-derived near-surface imagery against simpler approaches for various surface types, and perform a quantitative evaluation. Existing mapped daily imagery from the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (SNPP) and NOAA-20, and the Ocean and Land Colour Instrument (OLCI) on the Sentinel-3A and Sentinel-3B satellites is used to produce global clear sky near-surface imagery over various time intervals. We provide several examples of satellite-derived clear sky near-surface imagery over various regions to show potential applications. In addition, we apply this new method to derive clear sky near-surface imagery using higher spatial resolution Landsat-8 data, and discuss characteristics and limitations of our approach. The clear sky near-surface imagery is a useful satellite-derived product, representing the human perception of Earth’s near-surface features, which can be more directly interpreted and easily understood by the general public, and aids visualization and interpretation of various types of satellite derived data.

Effectiveness of machine learning methods for water segmentation with ROI as the label: A case study of the Tuul River in Mongolia

The carrying capacity of water resources is key to the sustainable development of arid and semi-arid regions. There are important challenges related to the detection of discontinuous and crooked water bodies in the vast Mongolian Plateau, despite the availability of remote sensing technology which has the advantage of facilitating water observations over large areas and timelines. Given the high cost and low coverage of high-resolution images and the low resolution of images with high coverage, this study proposes a pixel-based convolutional neural network (CNN) method for the application of water extracted from the region of interest (ROI) to medium-resolution Landsat images. The pixel-based CNN method combines the texture and spectral features of the ground object by connecting the center pixels of the images to the surrounding pixels. ROI is used instead of full-label datasets, reduce the difficulty of building labels in low-to-medium-resolution images. Taking the Tuul River in Mongolia as a case, the pixel-based CNN method, the normalized difference water index threshold (NDWI) method, the modified normalized difference water index (MNDWI) threshold method, U-net model in deep learning, and the pixel-based deep neural network (DNN) method were used with medium-resolution Landsat 8 images with ROI labels. The pixel-based CNN method shows better water extraction results for the cloud, cloud shadows, and building areas, compared with other methods. The method proposed in this study had the highest verification accuracy (92.07%). It also has the advantages of fewer training parameters and shorter training time. The training accuracies of the pixel-based CNN, pixel-based DNN, and U-net were 99.90%, 96.98%, and 93.70%, respectively. All training models and calling methods were uploaded to GitHub (https://github.com/CaryLee17/Pixel-based-CNN).

Satellite remote sensing of pelagic Sargassum macroalgae: The power of high resolution and deep learning

In recent years, massive blooms of pelagic Sargassum have occurred in the Atlantic Ocean, Caribbean Sea, and Gulf of Mexico, and satellite imagery have been used operationally to monitor and track the blooms. However, limited by the coarse resolution and other confounding factors, there is often a data gap in nearshore waters, and the uncertainties in the estimated Sargassum abundance in offshore waters are also unclear. Higher-resolution satellite data may overcome these limitations, yet such a potential is hindered by the lack of reliable methods to accurately detect and quantify Sargassum in an automatic fashion. Here, we address this challenge by combining large quantities of high-resolution satellite data with deep learning. Specifically, data from the Multispectral Instrument (MSI, 10–20 m), Operational Land Imager (OLI, 30 m), WorldView-II (WV-2, 2 m), and PlanetScope/Dove (3 m) are used with a deep convolution neural network (DCNN) to extract Sargassum features and quantify Sargassum biomass density or areal coverage. By utilizing the U-net architecture and the pre-trained weights from the VGG16 model, the DCNN (i.e., the VGGUnet model) can extract Sargassum features while discarding other confusing features (waves, currents, phytoplankton blooms, clouds, cloud shadows, or striping noise). For Sargassum biomass estimated from OLI and MSI images, results indicate an accuracy of ~92% and 90%, respectively, when evaluated using images from the same sensor. When Sargassum areal coverage is estimated from WV-2 and Dove images, there is an accuracy of ~98% and 82%, respectively. When different sensors are cross-compared, OLI reveals ~30% more Sargassum biomass than MODIS from 14 OLI images collected in the Caribbean Sea (path/row: 001/050) for their commonly viewed observable areas, and ~ 180% more Sargassum biomass than MSI (N = 15, path/row: 001/050); such differences appear systematic (R2 = 0.98 and 0.73, respectively). Compared to the quasi-simultaneous MSI, OLI, and MODIS images, Dove shows higher Sargassum coverage. Higher-resolution sensors tend to observe more Sargassum because they can detect smaller-scale features that are missed by the coarser-resolution sensors, although the difference varies with time and location. The morphological characteristics of Sargassum features from these high-resolution data are also reported to facilitate management actions. The findings here not only fill the knowledge gaps and coverage gaps from previous studies, but more importantly pave the road toward operational monitoring and tracking Sargassum features in nearshore waters.

System-Independent Irradiance Sensorless ANN-Based MPPT for Photovoltaic Systems in Electric Vehicles

The integration of photovoltaic systems (PVS) in electric vehicles (EV) increases the vehicle’s autonomy by providing an additional energy source other than the battery. However, current solar cell technology generates around 200 W for a 1.4 m2 panel (to be installed on the roof of the EV) at stable irradiance conditions. This limitation in production and the sudden changes in irradiance produced by shadows of clouds, buildings, and other structures make developing a fast and efficient maximum power point tracking (MPPT) technique in this area necessary. This article proposes an artificial neural network (ANN)-based MPPT, called DS-ANN, that uses manufacturer datasheet parameters as inputs to the network to address this problem. The Bayesian backpropagation-regularization performs the training, ensuring that the MPPT technique operates satisfactorily on different PVS without retraining. We simulated the response of 20 commercial modules against actual irradiance data to validate the proposed method. The results show that our method achieves an average tracking efficiency of 99.66%, improving by 1.21% over an enhanced P&O method.

Cloud shadows drive vertical migrations of deep-dwelling marine life.

Many zooplankton and fishes vertically migrate on a diel cycle to avoid predation, moving from their daytime residence in darker, deep waters to prey-rich surface waters to feed at dusk and returning to depth before dawn. Vertical migrations also occur in response to other processes that modify local light intensity, such as storms, eclipses, and full moons. We observed rapid, high-frequency migrations, spanning up to 60 m, of a diel vertically migrating acoustic scattering layer with a daytime depth of 300 m in the subpolar Northeastern Pacific Ocean. The depth of the layer was significantly correlated, with an ∼5-min lag, to cloud-driven variability in surface photosynthetically available radiation. A model of isolume-following swimming behavior reproduces the observed layer depth and suggests that the high-frequency migration is a phototactic response to absolute light level. Overall, the cumulative distance traveled per day in response to clouds was at least 36% of the round-trip diel migration distance. This previously undescribed phenomenon has implications for the metabolic requirements of migrating animals while at depth and highlights the powerful evolutionary adaptation for visual predator avoidance.

Automatic cloud and cloud shadow detection in tropical areas for PlanetScope satellite images

PlanetScope satellite data with a 3-m resolution and near-daily global coverage have been increasingly used for land surface monitoring, ranging from land cover change detection to vegetative biophysics characterization and ecological assessments. Similar to other satellite data, effective screening of clouds and cloud shadows in PlanetScope images is a prerequisite for these applications, yet remains challenging as PlanetScope has 1) fewer spectral bands than other satellites hindering the use of traditional methods, and 2) inconsistent radiometric calibration across satellite sensors making the cloud/shadow detection using fixed thresholds unrealistic. To address these challenges, we developed a SpatioTemporal Integration approach for Automatic Cloud and Shadow Screening (‘STI-ACSS’), including two steps: (1) generating initial masks of clouds/shadows by integrating both spatial (i.e. cloud/shadow indices of an individual PlanetScope image) and temporal (i.e. reflectance outliers in PlanetScope image time series) information with an adaptive threshold approach; (2) a two-step fine-tuning on these initial masks to derive final masks by integrating morphological processing with an object-based cloud and cloud shadow matching. We tested STI-ACSS at six tropical sites representative of different land cover types (e.g. forest, urban, cropland, savannah, and shrubland). For each site, we evaluated the performance of STI-ACSS with reference to the manual masks of clouds/shadows, and compared it with four state-of-the-art methods, namely Function of mask (Fmask), Automatic Time-Series Analysis (ATSA), Iterative Haze Optimized Transformation (IHOT) and the default PlanetScope quality control layer. Our results show that, across all sites, STI-ACSS 1) has the highest average overall accuracy (98.03%), 2) generates an average producer accuracy of 95.53% for clouds and 89.48% for cloud shadows, and 3) is robust across sites and seasons. These results suggest the effectiveness of using STI-ACSS for cloud/shadow detection for PlanetScope satellites in the tropics, with potential to be extended to other satellite sensors with limited spectral bands.

Evaluation of Five Deep Learning Models for Crop Type Mapping Using Sentinel-2 Time Series Images with Missing Information

Accurate crop type maps play an important role in food security due to their widespread applicability. Optical time series data (TSD) have proven to be significant for crop type mapping. However, filling in missing information due to clouds in optical imagery is always needed, which will increase the workload and the risk of error transmission, especially for imagery with high spatial resolution. The development of optical imagery with high temporal and spatial resolution and the emergence of deep learning algorithms provide solutions to this problem. Although the one-dimensional convolutional neural network (1D CNN), long short-term memory (LSTM), and gate recurrent unit (GRU) models have been used to classify crop types in previous studies, their ability to identify crop types using optical TSD with missing information needs to be further explored due to their different mechanisms for handling invalid values in TSD. In this research, we designed two groups of experiments to explore the performances and characteristics of the 1D CNN, LSTM, GRU, LSTM-CNN, and GRU-CNN models for crop type mapping using unfilled Sentinel-2 (Sentinel-2) TSD and to discover the differences between unfilled and filled Sentinel-2 TSD based on the same algorithm. A case study was conducted in Hengshui City, China, of which 70.3% is farmland. The results showed that the 1D CNN, LSTM-CNN, and GRU-CNN models achieved acceptable classification accuracies (above 85%) using unfilled TSD, even though the total missing rate of the sample values was 43.5%; these accuracies were higher and more stable than those obtained using filled TSD. Furthermore, the models recalled more samples on crop types with small parcels when using unfilled TSD. Although LSTM and GRU models did not attain accuracies as high as the other three models using unfilled TSD, their results were almost close to those with filled TSD. This research showed that crop types could be identified by deep learning features in Sentinel-2 dense time series images with missing information due to clouds or cloud shadows randomly, which avoided spending a lot of time on missing information reconstruction.

EVALUATION OF SAR TO OPTICAL IMAGE TRANSLATION USING CONDITIONAL GENERATIVE ADVERSARIAL NETWORK FOR CLOUD REMOVAL IN A CROP DATASET

Abstract. Most methods developed to map crop fields with high-quality are based on optical image time-series. However, often accuracy of these approaches is deteriorated due to clouds and cloud shadows, which can decrease the availably of optical data required to represent crop phenological stages. In this sense, the objective of this study was to implement and evaluate the conditional Generative Adversarial Network (cGAN) that has been indicated as a potential tool to address the cloud and cloud shadow removal; we also compared it with the Witthaker Smother (WS), which is a well-known data cleaning algorithm. The dataset used to train and assess the methods was the Luis Eduardo Magalhães benchmark for tropical agricultural remote sensing applications. We selected one MSI/Sentinel-2 and C-SAR/Sentinel-1 image pair taken in days as close as possible. A total of 5000 image pair patches were generated to train the cGAN model, which was used to derive synthetic optical pixels for a testing area. Visual analysis, spectral behaviour comparison, and classification were used to evaluate and compare the pixels generated with the cGAN and WS against the pixel values from the real image. The cGAN provided consistent pixel values for most crop types in comparison to the real pixel values and outperformed the WS significantly. The results indicated that the cGAN has potential to fill cloud and cloud shadow gaps in optical image time-series.

Multi-spectral cloud detection based on a multi-dimensional and multi-grained dense cascade forest

Cloud detection in satellite images is a vital step for cloud/land recognition, cloud/snow discrimination, and cloud shadow removal. Accurate cloud detection plays an important role in land resource management, environmental pollution monitoring, and land target recognition. Deep learning (DL) algorithms have shown great progress in cloud detection. However, as the complexity of the DL-based model increases, cloud detection efficiency decreases. DL-based cloud detection models are unable to successfully balance the performance-efficiency tradeoff. In our study, a multi-dimensional and multi-grained dense cascade forest (MDForest) is proposed for multi-spectral cloud detection. MDForest is a deep forest structure that automatically extracts low-level and high-level features from satellite cloud images end-to-end; a multi-dimensional and multi-grained scanning mechanism is introduced to capture the spectral information of multi-spectral satellite images while enhancing the representation learning ability of cascade forest. The experimental results on the HJ-1A/1B dataset show that MDForest improves the performance of cloud detection and possesses a good inference efficiency compared with DL-based cloud detection methods, which makes the proposed MDForest satisfy the application where good performance and high efficiency are both required.

GIS-Based Distribution System Planning for New PV Installations

Solar panel installations have increased significantly in Japan in recent decades. Due to this, world trends, such as clean/renewable energy, are being implemented in power systems all across Japan—particularly installations of photovoltaic (PV) panels in general households. In this work, solar power was estimated using solar radiation data from geographic information system (GIS) technology. The solar power estimation was applied to the actual distribution system model of the Jono area in Kitakyushu city, Japan. In this work, real power consumption data was applied to a real world distribution system model. We studied the impact of high installation rates of solar panels in Japanese residential areas. Additionally, we considered the voltage fluctuations in the distribution system model by assessing the impact of cloud shadows using a novel cloud movement simulation algorithm that uses real world GIS data. The simulation results revealed that the shadow from the cloud movement process directly impacted the solar power generation in residential areas, which caused voltage fluctuations of the overall distribution system. Thus, we advocate distribution system planning with a large number of solar panels.

Accurate extraction of surface water in complex environment based on Google Earth Engine and Sentinel-2.

To realize the accurate extraction of surface water in complex environment, this study takes Sri Lanka as the study area owing to the complex geography and various types of water bodies. Based on Google Earth engine and Sentinel-2 images, an automatic water extraction model in complex environment(AWECE) was developed. The accuracy of water extraction by AWECE, NDWI, MNDWI and the revised version of multi-spectral water index (MuWI-R) models was evaluated from visual interpretation and quantitative analysis. The results show that the AWECE model could significantly improve the accuracy of water extraction in complex environment, with an overall accuracy of 97.16%, and an extremely low omission error (0.74%) and commission error (2.35%). The AEWCE model could effectively avoid the influence of cloud shadow, mountain shadow and paddy soil on water extraction accuracy. The model can be widely applied in cloudy, mountainous and other areas with complex environments, which has important practical significance for water resources investigation, monitoring and protection.

Recognition and counting of pitaya trees in karst mountain environment based on unmanned aerial vehicle RGB images

The counting of crop plants is an important basis for the estimation of pitaya yield. Traditional crop monitoring methods are time-consuming and laborious. The pitaya tree, one of the characteristic economic crops in the complex mountain environment, was selected as the research object. Considering the comprehensive factors such as different seasons, cloud shadow shading, crop interplanting, steep terrain, different breeds, and ages of pitaya trees, the quad-rotor unmanned aerial vehicles (UAVs) were used to collect the visible light images of the planting base of pitaya trees in the test area in karst plateau canyon region and the verification area in a complex environment. First, the characteristics of the RGB values of the eight main ground objects in the test areas are analyzed according to the color index, geometric size, and texture characteristics, which show that pitaya plants, weeds, and shrubs have certain interference with each other. Then the excess green index (ExG) of the images is calculated to enhance the vegetation characteristics and separate vegetation and non-vegetation. Gaussian high-pass filter (GHPF) is used to retain the high-frequency information of pitaya plants on the ExG images. After GHPF processing, the DN values of shrub and weeds are reduced, the edges of the area target are enhanced, and the influences of weeds and cloud shadows on plant identification are eliminated. Finally, through field measurement of pitaya plant data and OTSU, grayscale segmentation was performed on the images processed by GHPF and the pitaya plant information was extracted. Combined with the projection area of single pitaya tree, the number of pitaya trees was obtained by area screening method. The target detection percentage of test area A, test area B, and the verification area in complex environment is, respectively, 96.99%, 94.66%, and 94.30%; the quality percentage of them is, respectively, 92.46%, 90.41% and 91.50%; the branching factor of them is, respectively, 0.05, 0.05, and 0.03, proving that the RGB images collected by UAV can be used in the recognition and counting of pitaya trees in complex mountain environments. Our study can provide a reference for the application of low-cost, high-efficiency UAV visible light remote sensing to precision agriculture in complex mountain environment.