High Resolution Multispectral Data Research Articles

TreeSeg—A Toolbox for Fully Automated Tree Crown Segmentation Based on High-Resolution Multispectral UAV Data

Single-tree segmentation on multispectral UAV images shows significant potential for effective forest management such as automating forest inventories or detecting damage and diseases when using an additional classifier. We propose an automated workflow for segmentation on high-resolution data and provide our trained models in a Toolbox for ArcGIS Pro on our GitHub repository for other researchers. The database used for this study consists of multispectral UAV data (RGB, NIR and red edge bands) of a forest area in Germany consisting of a mix of tree species consisting of five deciduous trees and three conifer tree species in the matured closed canopy stage at approximately 90 years. Information of NIR and Red Edge bands are evaluated for tree segmentation using different vegetation indices (VIs) in comparison to only using RGB information. We trained Faster R-CNN, Mask R-CNN, TensorMask and SAM in several experiments and evaluated model performance on different data combinations. All models with the exception of SAM show good performance on our test data with the Faster R-CNN model trained on the red and green bands and the Normalized Difference Red Edge Index (NDRE) achieving best results with an F1-Score of 83.5% and an Intersection over Union of 65.3% on highly detailed labels. All models are provided in our TreeSeg toolbox and allow the user to apply the pre-trained models on new data.

Assessing the Impact of Clearing and Grazing on Fuel Management in a Mediterranean Oak Forest through Unmanned Aerial Vehicle Multispectral Data

Climate change has intensified the need for robust fire prevention strategies. Sustainable forest fuel management is crucial in mitigating the occurrence and rapid spread of forest fires. This study assessed the impact of vegetation clearing and/or grazing over a three-year period in the herbaceous and shrub parts of a Mediterranean oak forest. Using high-resolution multispectral data from an unmanned aerial vehicle (UAV), four flight surveys were conducted from 2019 (pre- and post-clearing) to 2021. These data were used to evaluate different scenarios: combined vegetation clearing and grazing, the individual application of each method, and a control scenario that was neither cleared nor purposely grazed. The UAV data allowed for the detailed monitoring of vegetation dynamics, enabling the classification into arboreal, shrubs, herbaceous, and soil categories. Grazing pressure was estimated through GPS collars on the sheep flock. Additionally, a good correlation (r = 0.91) was observed between UAV-derived vegetation volume estimates and field measurements. These practices proved to be efficient in fuel management, with cleared and grazed areas showing a lower vegetation regrowth, followed by areas only subjected to vegetation clearing. On the other hand, areas not subjected to any of these treatments presented rapid vegetation growth.

CloudS2Mask: A novel deep learning approach for improved cloud and cloud shadow masking in Sentinel-2 imagery

The Sentinel-2 satellite constellation produces high-resolution multispectral data, covering the entire Earth's land surface every five days. However, the use of this data is significantly impacted by the presence of pixels affected by cloud and shadow. Precise and efficient cloud and cloud shadow masking methods are required for the automated use of this data. Here, we present CloudS2Mask, a new and open-source Python deep-learning library to detect cloud and cloud shadows in Sentinel-2 data accurately and efficiently. CloudS2Mask is trained on multiple large global datasets including CloudSEN12, KappaSet, Sentinel-2 Cloud Mask Catalogue, Sentinel-2 reference cloud masks generated by an active learning method and a custom CloudS2Mask dataset. Using techniques such as progressive resizing, hard fine-tuning, and transfer learning, CloudS2Mask outperforms published models when evaluated on both the CloudSEN12 and the PixBox datasets in terms of classification accuracy. Compared to the next best-published results, CloudS2Mask improves Balanced Overall Accuracy metrics on the CloudSEN12 dataset by 1.2% for cloud, 2.8% for cloud shadow and 1.1% for clear classification. CloudS2Mask provides a range of customisable options combined with novel pre-processing and post-processing approaches to meet task-specific needs, enabling mean scene processing speeds up to 44 times faster than current state-of-the-art methods. The application of novel techniques in CloudS2Mask, combined with its accuracy and processing speed, underscores its potential to streamline the automated cloud and cloud shadow detection of Sentinel-2 data.

Water Quality Estimation Using Gaofen-2 Images Based on UAV Multispectral Data Modeling in Qinba Rugged Terrain Area

This study presents an innovative method for large-scale surface water quality assessment in rugged terrain areas, specifically tailored for regions like the Qinba Mountains. The approach combines the use of high-resolution (10 cm) multispectral data acquired by unmanned aerial vehicles (UAVs) with synchronized ground sampling and 1 m resolution multispectral imagery from China’s Gaofen-2 satellite. By integrating these technologies, the study aims to capitalize on the convenience and synchronized observation capabilities of UAV remote sensing, while leveraging the broad coverage of satellite remote sensing to overcome the limitations of each individual technique. Initially, a multispectral estimation model is developed for key water quality parameters, including chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP), utilizing data from UAVs and coordinated ground samples. Subsequently, a comparison is made between the spectral band ratios (R/G and NIR/G) obtained from the UAV data and those from the Gaofen-2 satellite data, revealing a substantial similarity. Ultimately, this integrated methodology is successfully employed in monitoring water quality across a vast area, particularly along the midstream of the Hanjiang River in the Qinba Mountain region. The results underscore the feasibility, advantages, improved efficiency, and enhanced accuracy of this approach, making it particularly suitable for large-scale water quality monitoring in mountainous terrain. Furthermore, this method reduces the burden associated with traditional ground-based spectral acquisitions, paving the way for a more practical and cost-effective solution in monitoring vast water bodies.

Evaluating Different Deep Learning Approaches for Tree Health Classification Using High-Resolution Multispectral UAV Data in the Black Forest, Harz Region, and Göttinger Forest

We present an evaluation of different deep learning and machine learning approaches for tree health classification in the Black Forest, the Harz Mountains, and the Göttinger Forest on a unique, highly accurate tree-level dataset. The multispectral UAV data were collected from eight forest plots with diverse tree species, mostly conifers. As ground truth data (GTD), nearly 1500 tree polygons with related attribute information on the health status of the trees were used. This data were collected during extensive fieldwork using a mobile application and subsequent individual tree segmentation. Extensive preprocessing included normalization, NDVI calculations, data augmentation to deal with the underrepresented classes, and splitting the data into training, validation, and test sets. We conducted several experiments using a classical machine learning approach (random forests), as well as different convolutional neural networks (CNNs)—ResNet50, ResNet101, VGG16, and Inception-v3—on different datasets and classes to evaluate the potential of these algorithms for tree health classification. Our first experiment was a binary classifier of healthy and damaged trees, which did not consider the degree of damage or tree species. The best results of a 0.99 test accuracy and an F1 score of 0.99 were obtained with ResNet50 on four band composites using the red, green, blue, and infrared bands (RGBI images), while VGG16 had the worst performance, with an F1 score of only 0.78. In a second experiment, we also distinguished between coniferous and deciduous trees. The F1 scores ranged from 0.62 to 0.99, with the highest results obtained using ResNet101 on derived vegetation indices using the red edge band of the camera (NDVIre images). Finally, in a third experiment, we aimed at evaluating the degree of damage: healthy, slightly damaged, and medium or heavily damaged trees. Again, ResNet101 had the best performance, this time on RGBI images with a test accuracy of 0.98 and an average F1 score of 0.97. These results highlight the potential of CNNs to handle high-resolution multispectral UAV data for the early detection of damaged trees when good training data are available.

Lithological mapping of Nidar ophiolite complex, Ladakh using high-resolution data

Ophiolites represents a distinctive variety of igneous rock that comprises fragments of oceanic crust with the underlying mantle portion. Their mineral assemblages and distribution relation, both spatial and temporal are in close association with the complex mantle dynamics of the Earth. Nidar ophiolite complex located towards SE of Ladakh is one of such ophiolite assemblages exposed well in the Indus Suture Zone of Himalayas. It represents parts of Neo-Tethys ocean that existed during Mesozoic-Cenozoic period between continental plates of India and Eurasia. The earlier studies for Indian ophiolites are mostly field and analytical based with a limited work carried out using remote sensing techniques and that adds scope to apply this approach for further exploration of ophiolites. Advanced tools and techniques especially, hyperspectral remote sensing and machine learning based approach can be effective operational tools for exploration of ophiolites that provide assessable terrains for characterization of Earth’s mantle, serve analogs for planetary crustal rocks and also have economic significance. This study has utilized high-resolution multispectral (Landsat OLI and ASTER) and hyperspectral data from various missions including the recent PRISMA mission which has provided the first hyperspectral datasets for lithological mapping of Nidar ophiolite section. The multiple high-resolution datasets were used for identification and mapping of the various lithounits of the Nidar ophiolites together with the underlying and overlying associated rocks based on band ratioing, derived spectral compositional indices and spectral characterization. The differentiated lithounits of the Nidar ophiolites are interpreted based on spectral response of lithological units in different bands. In addition, classification based on machine learning techniques have been applied to the PRISMA data to generate a detailed classification map of the study area where the various lithological units of the ophiolites and associated rock exposures belonging to various geological formations are clearly discernible. The results have been validated from geological field visit to the study area and spectral analysis of the field samples.

Forest Gap Extraction Based on Convolutional Neural Networks and Sentinel-2 Images

As a type of small-scale disturbance, forest gap and its accurate extraction are of great significance to monitor forest long-term dynamics, to choose forest recovery mode and to predict forest recovery pace. Currently, airborne LiDAR and high-resolution multi-spectral data are commonly used to accurately classify forest gaps, but they are costly to acquire and have limited time and space availability. In contrast, the Sentinel-2 multi-spectral data with a 10 m spatial resolution overcomes these drawbacks in forest gap extraction. In this work, an integrated framework that combines multi-source remote sensing, machine learning and deep learning to extract forest gap in wide regions was proposed and tested in three sites. First, LiDAR, Sentinel series and random forest (RF) algorithm were synergized to produce a canopy height product in model training site. On this basis, samples for forest canopy, forest gap and non-such were identified from LiDAR-derived canopy height model (CHM) and Sentinel-based canopy height inversion (HI) data to train forest gap extraction models by applying the Deep Forest (DF) and Convolutional Neural Networks (CNN) algorithms, followed by a comparison of the accuracy and the transferability among the four models (DF-CHM, DF-HI, CNN-CHM and CNN-HI). The results indicated that the R2 and RMSE of Sentinel-based canopy height retrievals were estimated at 0.63, and 7.85 m respectively, the difference in the mean height and standard deviation between HI and CHM was 0.03 m and 4.7 m respectively. And there was a spatial agreement of about 98.60% between the HI-identified samples and the CHM-identified samples, with an agreement of 54.89% for the forest gap class. The CNN-HI model had the highest accuracy in both transfer learning test sites, with an overall accuracy (OA) of 0.85 and 0.87, Kappa coefficient at 0.78 and 0.81, respectively, proving that it has good transferability. Conversely, the DF-based models generally gave poorer accuracy and transferability. This study demonstrates that combining Sentinel-2 multi-spectral data and CNN algorithm is feasible and effective in forest gap extraction applications over wide regions.

Design of Ground Target Based on RedEdge MX-Dual Multispectral Camera

Ground target plays an indispensable role in UAV near-earth remote sensing geometric positioning, geographic registration, correction and improvement of remote sensing image accuracy and accuracy. Traditional ground targets are markers with distinct features, shapes and textures, and their dimensions are usually fixed. The UAV can provide high-resolution multi-spectral data with multi-spectral camera, so the target resolution will change as the UAV height increases and the camera band increases, in some bands of the multi-spectral camera, there will be non-response, which will affect the accuracy of target recognition and positioning. Based on the color, material, size and shape of target, a set of targets suitable for target recognition and location of multi-spectral camera is designed in this paper. In order to judge whether the target performance meets the requirements, this paper verifies the positioning accuracy of the target and its response in each channel of the camera. It is proved that the target plane error is about 10mm at 20 ~ 100m altitude, and the height error is about 40mm, and it can respond in ten channels, it provides experience and ideas for the design and development of multi-spectral camera targets.

Integrating Sentinel 2A data for selecting the best methods in observation of biological pigments in coastal water

Coastal waters serve as critical ecosystems, supporting diverse marine life and offering invaluable resources to human communities. Understanding the distribution and dynamics of biological pigments, such as chlorophyll and phytoplankton, is essential for assessing the health and productivity of these ecosystems. Remote sensing technology has revolutionized our ability to monitor these pigments, providing a synoptic view of coastal waters. Sentinel-2A, a European Space Agency satellite, offers high-resolution multispectral data, making it a valuable resource for coastal studies. This paper presents a comprehensive analysis of Sentinel-2A data integration techniques for the observation of biological pigments in coastal waters along the coasts of Mediterranean Sea (Abu-Qir Area) from seven different sampling sites for the presence of different chlorophyll forms. Water samples were collected and physico-chemical characteristics like dissolved oxygen, pH, conductivity and chlorophyll were measured. In chlorophyll estimation, different forms of were calculated viz. chl a, b, c and carotenoids using different methodologies. Images from the Sentinel Multispectral Imager (MSI) sensor were taken on the same sample date. The phyto-planktonic forms, carotenoids, and their ratios in the sampling sites were also determined analytically. With the integration of satellite images (Sentinel 2A), results indicated types of chlorophyll, carotenoids, and ratios' distributions in the Abu-Qir Area of the Mediterranean Sea, Egypt. Our results demonstrate that the kind of species, illumination levels, and nutrient concentrations may all affect how much chlorophyll is absorbed by phytoplankton. Additionally, the band ratios of the different pigment types were found to be connected, such as b3/b8, or the same band might interact with other bands like b5, b4, b6, and b7, or b8a. Additionally, it was noticed that the green band and near infrared bands were useful for detecting pigments integrated with sampling datasets.

Three-Dimensional Mapping of Habitats Using Remote-Sensing Data and Machine-Learning Algorithms

Progress toward habitat protection goals can effectively be performed using satellite imagery and machine-learning (ML) models at various spatial and temporal scales. In this regard, habitat types and landscape structures can be discriminated against using remote-sensing (RS) datasets. However, most existing research in three-dimensional (3D) habitat mapping primarily relies on same/cross-sensor features like features derived from multibeam Light Detection And Ranging (LiDAR), hydrographic LiDAR, and aerial images, often overlooking the potential benefits of considering multi-sensor data integration. To address this gap, this study introduced a novel approach to creating 3D habitat maps by using high-resolution multispectral images and a LiDAR-derived Digital Surface Model (DSM) coupled with an object-based Random Forest (RF) algorithm. LiDAR-derived products were also used to improve the accuracy of the habitat classification, especially for the habitat classes with similar spectral characteristics but different heights. Two study areas in the United Kingdom (UK) were chosen to explore the accuracy of the developed models. The overall accuracies for the two mentioned study areas were high (91% and 82%), which is indicative of the high potential of the developed RS method for 3D habitat mapping. Overall, it was observed that a combination of high-resolution multispectral imagery and LiDAR data could help the separation of different habitat types and provide reliable 3D information.

Mapping intertidal oyster farms using unmanned aerial vehicles (UAV) high-resolution multispectral data

In France, oyster aquaculture has been historically developed in intertidal zones, with shellfish farming areas covering much of the Atlantic coast. Monitoring these off-bottom cultures where oysters are grown in plastic mesh-bags set on trestle tables is mandatory for the maritime administration to check compliance with a Structural Plan Document (SPD), while also being important for stock assessment in relation to carrying capacity issues. However, traditional monitoring methods are time-consuming, labor-intensive, and inefficient in covering large intertidal areas. In this study, we used a new GIS-based analytical method to assess the potential of high-resolution Unmanned Aerial Vehicle (UAV) multispectral data to retrieve spatial information on oyster-farming structures using Bourgneuf Bay (France) as a case-study. A non-parametric machine learning algorithm was applied to four UAV flight orthomosaics collected at different altitudes (12, 30, 50, and 120 m) to identify oyster mesh-bags. These supervised classifications achieved overall accuracies above 95% for all tested altitudes. In addition, an accurate distinction of oyster-bag mesh sizes (4, 9 and 14 mm) was obtained for 12–50 m flights, but there was a lower accuracy at 120 m. Across all flights, the 4 mm mesh size was the least well detected (72.14% Producer Accuracy). This information can be used to identify bags with specific mesh-sizes used for spat or adult grow out. Finally, we accurately measured oyster table heights using a high-resolution Digital Surface Model (DSM) derived from Structure from Motion (SfM) photogrammetry. The 50 m flight was suggested as the best compromise to obtain precise measurements of the oyster table heights while covering larger areas than lower altitude flights. This demonstrates that UAV technology can provide a set of spatial variables relevant for shellfish farmers and coastal managers in an efficient, rapid, and non-destructive way to monitor the extent and characteristics of oyster-farming areas regularly.

Identification of Tree Species in Forest Communities at Different Altitudes Based on Multi-Source Aerial Remote Sensing Data

The accurate identification of forest tree species is important for forest resource management and investigation. Using single remote sensing data for tree species identification cannot quantify both vertical and horizontal structural characteristics of tree species, so the classification accuracy is limited. Therefore, this study explores the application value of combining airborne high-resolution multispectral imagery and LiDAR data to classify tree species in study areas of different altitudes. Three study areas with different altitudes in Muyu Town, Shennongjia Forest Area were selected. Based on the object-oriented method for image segmentation, multi-source remote sensing feature extraction was performed. The recursive feature elimination algorithm was used to filter out the feature variables that were optimal for classifying tree species in each altitude study area. Four machine learning algorithms, SVM, KNN, RF, and XGBoost, were combined to classify tree species at each altitude and evaluate the accuracy. The results show that the diversity of tree layers decreased with the altitude in the different study areas. The texture features and height features extracted from LiDAR data responded better to the forest community structure in the different study areas. Coniferous species showed better classification than broad-leaved species within the same study areas. The XGBoost classification algorithm showed the highest accuracy of 87.63% (kappa coefficient of 0.85), 88.24% (kappa coefficient of 0.86), and 84.03% (kappa coefficient of 0.81) for the three altitude study areas, respectively. The combination of multi-source remote sensing numbers with the feature filtering algorithm and the XGBoost algorithm enabled accurate forest tree species classification.

Almond cultivar identification using machine learning classifiers applied to UAV-based multispectral data

ABSTRACT In Portugal, almonds are a very important crop, due to their nutritional properties. In the northeastern part of the country, the almond sector has endured over time, with strong cultural traditions and key economic significance. In these areas, several cultivars are used. In effect, the presence of various almond cultivars implies differentiated management in irrigation, disease control, pruning system, and harvest planning. Therefore, cultivar classification is essential over large agricultural areas. Over the last decades, remote-sensing data have led to important breakthroughs in the classification of different cultivars for several crops. Nonetheless, for almonds, studies are incipient. Thus, this study aims to fill this knowledge gap and explore the classification of almond cultivars in an almond orchard. High-resolution multispectral data were acquired by an unmanned aerial vehicle (UAV). Vegetation indices (VIs) and tree structural parameters were, subsequently, estimated. To obtain an accurate cultivar identification, four machine learning classifiers, such as K-nearest neighbour (kNN), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost), were applied and optimized through the fine-tuning process. The accuracy of machine learning classifiers was analysed. SVM and RF performed best with OAs of 76% and 74% using VIs and spectral bands (GREEN, GRVI, GN, REN, ClRE). Adding the canopy height model (CHM) improved performance, with RF and XGBoost having OAs of 88% and 84%. kNN performed worst with an OA of 73% using only VIs and spectral bands, 80% with VIs, spectral bands and CHM, and 93% with VIs, CHM, and tree crown area (TCA). The best performance was achieved by RF and XGBoost with OAs of 99% using VIs, CHM, and TCA. These results demonstrate the importance of the feature selection process. Moreover, this study reveals the feasibility of remote-sensing data and machine learning classifiers in the classification of almond cultivars.

Atlantic forest woody carbon stock estimation for different successional stages using Sentinel-2 data

The Atlantic Forest is one of the most threatened biodiversity hotspots and environmental impacts has made its landscape fragmented and heterogeneous. The heterogeneity of the fragments is a challenge for the characterization and quantification of forest resources, such as the stock of biomass and carbon. Methodologies based on remote sensing have been used, to improve these estimates without compromising execution costs. The objective was to estimate, with high spatial resolution passive remote sensing, the aboveground carbon stock in fragments of different successional stages of the Atlantic Forest. Forests were classified into initial, intermediate, and advanced successional stages. In each stratum, 10 plots (20x50 m) were established, and the carbon stock was calculated by adjusted Schumacher and Hall model. The reflectances of the blue, green, red, and near-infrared bands and vegetation indices (VIs) were obtained in the dry and rainy seasons, from MSI/Sentinel-2 images, with a resolution of 10 m. Artificial Neural Networks (ANN), with different combinations of variables, were trained and validated with simulated reflectance values. Carbon was estimated by ANN with the best performance in training and validation. The average carbon stock in the initial, intermediate, and advanced strata was 24.99, 35.79 and 82.28 Mg ha−1, respectively, with a general average of 47.68 Mg ha−1. The carbon estimates were better with the ANN trained with the reflectances of the rainy season. The addition of VIs did not improve ANN performance. The simulated spectral data were consistent and adequate to validate the selected ANN. The total carbon stock, modeled was 41,962.15 Mg, ranging from 6.68 to 108.29 Mg ha−1, with an average of 48.70 Mg ha−1. The carbon stock in the advanced stratum is more than three times that observed in the initial stratum, and they were efficiently estimated using high-resolution multispectral data, obtained in the rainy season, as inputs.

Multiview Learning for Impervious Surface Mapping Using High-Resolution Multispectral Imagery and LiDAR Data

The use of multi-source remote sensing data to obtain urban impervious surface has become a popular research topic. Multi-source remote sensing data fusion techniques can provide object interpretation with a higher accuracy. However, most decision-level fusion methods make insufficient use of the complementary information and degree of association between similar object data. To fill this gap, in this paper, we propose a dual-view learning fusion classification method (DvLF) based on multi-view learning. First, DvLF uses co-training algorithm to combine multiple data sources for accurate classification, extracting easy-to-classify area while separating difficult-to-classify regions for further analysis. Secondly, a canonical correlation analysis method is adopted to mine the degree of association of similar object data for constructing a subspace projection field of each object sample. The data in the difficult-to-classify regions are classified in the projection field of each object, and then the results of each classification are fused by voting. Finally, the classification results of the two regions are combined into the classification results of the whole image to achieve impervious surface mapping. The proposed method is applied to the dual-sensor (high-resolution image and LiDAR) Buffalo dataset and the dual-sensor (RGB and multispectral LiDAR) Houston dataset. The experimental results show that our method achieved a significant improvement in classification accuracy compared to other methods. The overall classification accuracy of this new DvLF fusion method on the Buffalo and Houston datasets is 83.35% and 88.84%, respectively leading to accurate high-resolution impervious surface mapping.

Integration of C-band SAR and high-resolution optical images for delineating palaeo-channels in Nagaur and Barmer districts, western Rajasthan, India.

Identifying hitherto unknown palaeo-channels, especially in the arid regions of the Thar Desert, is crucial since these channels may form excellent aquifers, and are also associated with valuable ore deposits of many precious minerals. This study employed integrated C-band Synthetic Aperture Radar (SAR) of Sentinel-1A and high-resolution multispectral Sentinel-2A data of pre- and post-monsoon seasons (June and November) to delineate playas and palaeo-channels. This approach is the first of its kind for this area. The palaeo-channels were delineated through a detailed visual inspection of colour composite (CC) images of Sentinel-2A data, SAR backscatter (VH) images and fused SAR and optical images. Then, we studied the topographic profiles generated from the Shuttle Radar Topography Mission -Digital Elevation Model (SRTM-DEM) across the identified palaeo-channels, Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI) to further confirm the existence of a palaeo-channel's course and playas. As a result, several playas and palaeo-channels in the area were successfully identified, some of which were previously unmapped and undetected. The results indicate that the post-monsoon datasets are more useful for the precise delineation of palaeo-channels due to the presence of relatively higher moisture along the palaeo-channels' courses.

Comparison of UAV and SAR performance for Crop type classification using machine learning algorithms: a case study of humid forest ecology experimental research site of West Africa

ABSTRACT Food insecurity is one of the major challenges facing African countries; therefore, timely and accurate information on agricultural production is essential to feed the growing population on the continent. A synergistic approach comprising a high-resolution multispectral UAV optical dataset and synthetic aperture radar (SAR) can help understand spectral features of target objects, especially with crop type identification. We conducted this work on the experimental plots using high spatial resolution multispectral UAV data (12 cm, re-sampled to 50 cm) in combination with the Sentinel 1C Synthetic Aperture Radar (SAR) dataset. We generated 11 agronomically relevent vegetation indices from the UAV multispectral image. Multiple combinations of the UAV datasets were analysed to assess the impact of canopy height model (CHM) on classification accuracy and to determine the optimum dataset (including spatial resolution) for the land cover classification. We also appraise the impact of variable spatial resolution on classification accuracy. A combination of VH and VV polarizations of Sentinel-1 SAR data was also analysed to classify the crop types while comparing its accuracy with the UAV-derived models. Our results show that model accuracy is improved- for all the data combination pairs, when CHM is added to the modelling. We also observed a decreasing trend in classification accuracy with respect to increasing spatial resolution. Generally, the support vector machine (SVM) classifier produced an overall accuracy of 94.78% and 81.72% for UAV and SAR datasets, respectively. In comparison, the random forest (RF) achieved an accuracy of 93.84% and 92.58%, for UAV and SAR datasets, respectively. The outputs from ground-based validation corroborate the results from model-based classification coupled with acceptable simple models’ agreement ratio (SMAR), exceeding 90% in some cases. The combined techniques can be useful in precision agriculture over small and large agricultural fields to support food security assessment and planning.

HOW CHALLENGING IS THE DISCRIMINATION OF FLOATING MATERIALS ON THE SEA SURFACE USING HIGH RESOLUTION MULTISPECTRAL SATELLITE DATA?

Abstract. Developing a remote sensing-based monitoring system for detecting marine plastics requires a thorough investigation of their discrimination possibilities from other floating objects. To this end, this study aims to explore the ability to discriminate marine debris from other floating materials and sea features using high-resolution multispectral satellite data. To perform our analysis, we utilized the open-access Marine Debris Archive (MARIDA), which contains several marine classes on Sentinel-2 (S2) data and benchmarks machine learning frameworks. In particular, we investigated well-established spectral indices, GrayLevel Co-occurrence Matrix (GLCM), Local Binary Pattern (LBP) texture and other spatial features at multiple scales. A Random Forest (RF) classifier was also applied for the classification procedure, and the spectral and spatial features which contributed the most were underlined. The quantitative and qualitative assessment indicated that the spectral information alone is insufficient to distinguish marine plastic from other floating materials which exhibit similar spectral behavior, such as vessels. However, a strong potential even for challenging discrimination tasks is presented when combined with spatial information. By further evaluating our results qualitatively, significant insights are gained, and specific combinations are proposed for challenging floating materials discrimination.

MLR-DBPFN: A Multi-Scale Low Rank Deep Back Projection Fusion Network for Anti-Noise Hyperspectral and Multispectral Image Fusion

Fusing low spatial resolution (LR) hyperspectral (HS) data and high spatial resolution (HR) multispectral (MS) data aims to obtain HR HS data. However, due to bad weather and the aging of sensor equipment, HS images usually contain a lot of noise, e.g., Gaussian noise, strip noise, and mixed noise, which would make the fused image have low quality. To solve this problem, we propose the multiscale low-rank deep back projection fusion network (MLR-DBPFN). First, HS and MS are superimposed, and multiscale spectral features of the stacked image are extracted through multiscale low-rank decomposition and convolution operation, which effectively removes noisy spectral features. Second, the upsampling and downsampling network mechanisms are used to extract the multiscale spatial features from each layer of spectral features. Finally, the multiscale spectral features and multiscale spatial features are combined for network training, and the weight of the noisy spectrum features is reduced through the network feedback mechanism, which suppresses the noisy spectrum and improves the noisy HS fusion performance. Experimental results on datasets of different noise demonstrate that MLR-DBPFN has superior spatial and spectral fidelity, comparative fusion quality, and robust antinoise performance compared with state-of-the-art methods.

Data-Driven Cloud Clustering via a Rotationally Invariant Autoencoder

Advanced satellite-born remote sensing instruments produce high-resolution multi-spectral data for much of the globe at a daily cadence. These datasets open up the possibility of improved understanding of cloud dynamics and feedback, which remain the biggest source of uncertainty in global climate model projections. As a step towards answering these questions, we describe an automated rotation-invariant cloud clustering (RICC) method that leverages deep learning autoencoder technology to organize cloud imagery within large datasets in an unsupervised fashion, free from assumptions about predefined classes. We describe both the design and implementation of this method and its evaluation, which uses a sequence of testing protocols to determine whether the resulting clusters: (1) are physically reasonable, (i.e., embody scientifically relevant distinctions); (2) capture information on spatial distributions, such as textures; (3) are cohesive and separable in latent space; and (4) are rotationally invariant, (i.e., insensitive to the orientation of an image). Results obtained when these evaluation protocols are applied to RICC outputs suggest that the resultant novel cloud clusters capture meaningful aspects of cloud physics, are appropriately spatially coherent, and are invariant to orientations of input images. Our results support the possibility of using an unsupervised data-driven approach for automated clustering and pattern discovery in cloud imagery.