Search

Mastering the changes in the planting area of crops are critical for formulating sustainable agricultural development strategies. The Landsat 9 satellite provides continuous observational data for crop mapping. It is necessary to evaluate the potential of Landsat 9 images for canola mapping. According to literatures, some canola indices have been proposed in relation to the spectral features of the yellow canola flower, such as the canola ratio index (CRI), normalised difference yellowness index (NDYI), canola index (CI), canola flower index (CFI), normalised difference rapeseed flower index (NRFI) and winter rapeseed index (WRI). However, it is unclear which index has more potential to extract canola based on the Landsat 9 images. Therefore, the potentials of these indices for canola mapping were quantitatively compared by using different classification methods including supervised and unsupervised classification methods. For supervised classification methods and the CFI, the overall accuracy of canola mapping was above 90% and the kappa coefficient was above 0.8. For unsupervised classification methods, CFI also performs best. It was demonstrated that the CFI outperformed the other five indices for canola mapping. It was also confirmed that Landsat 9 images and the CFI exhibit promising potential for canola mapping via quantitative comparison. [ABSTRACT FROM AUTHOR]

Monitoring vegetation response to valley‐scale floodplain restoration to evaluate effectiveness can be costly and time‐consuming. We used publicly available National Agriculture Imagery Program (NAIP) data and commonly used ArcGIS software to assess land cover change over time at five study sites located in semi‐arid environments of eastern Oregon and north‐central California. Accuracy assessments of our unsupervised classifications were used to evaluate effectiveness. Overall accuracy across sites and years ranged from 64.2% to 89.2% with mean and median accuracy of 79.1% and 80.6%, respectively. Further, we compared our classifications with high‐resolution uncrewed aerial systems (UAS)‐based data collected in the same timeframe. Restored areas classified as dense vegetation were within 4% of the UAS study, water was within 6%, and post‐restoration classifications of sparse vegetation and bare ground classes were within 6% and 4% of the UAS study, respectively. This comparison demonstrates that our unsupervised NAIP data classification of land cover change across entire valley‐scale restoration projects can be used to monitor riparian vegetation change over time as accurately as UAS‐based methods, but at lower cost. Additionally, our methods leverage existing fine‐resolution, pre‐restoration vegetation density data that were not collected as part of project planning. [ABSTRACT FROM AUTHOR]

Supervised polarimetric synthetic aperture radar (PolSAR) image classification demands a large amount of precisely labeled data. However, such data are difficult to obtain. Therefore, many unsupervised methods have been proposed for unsupervised PolSAR image classification. The classification maps of unsupervised methods contain many high-confidence samples. These samples, which are often ignored, can be used as supervisory information to improve classification performance on PolSAR images. This study proposes a new unsupervised PolSAR image classification framework. The framework combines high-confidence superpixel pseudo-labeled samples and semi-supervised classification methods. The experiments indicated that this framework could achieve higher-level effectiveness in unsupervised PolSAR image classification. First, superpixel segmentation was performed on PolSAR images, and the geometric centers of the superpixels were generated. Second, the classification maps of rotation-domain deep mutual information (RDDMI), an unsupervised PolSAR image classification method, were used as the pseudo-labels of the central points of the superpixels. Finally, the unlabeled samples and the high-confidence pseudo-labeled samples were used to train an excellent semi-supervised method, similarity matching (SimMatch). Experiments on three real PolSAR datasets illustrated that, compared with the excellent RDDMI, the accuracy of the proposed method was increased by 1.70%, 0.99%, and 0.8%. The proposed framework provides significant performance improvements and is an efficient method for improving unsupervised PolSAR image classification. [ABSTRACT FROM AUTHOR]

Coastal wetland has many ecological service functions such as shoreline protection, biodiversity conservation, material production, energy exchange, and providing leisure and scientific research space. Vegetation is an important part of coastal wetland, and its distribution, structural changes and other landscape information reflect the health status of coastal wetlands to a large extent. In order to analyze the feasibility of identifying vegetation species under ebb and flow of coastal wetland by remote sensing, two periods of Landsat 8 ebb and flow images were selected to classify vegetation in coastal wetland of Yellow River Delta based on statistical discriminant, decision tree supervised classification and unsupervised classification methods. The results showed that the classification effect of statistical discriminant supervised classification was the best, the classification accuracy was up to 97%, and the vegetation types could be accurately distinguished, and the ebb and flow had little effect on the classification results. The research indicated that remote sensing technology is feasible to extract vegetation information of coastal wetland under different tidal conditions, which can provide technical and data support for vegetation monitoring, ecological restoration and blue carbon storage estimation of coastal wetland. [ABSTRACT FROM AUTHOR]

Mastering the changes in the planting area of crops are critical for formulating sustainable agricultural development strategies. The Landsat 9 satellite provides continuous observational data for crop mapping. It is necessary to evaluate the potential of Landsat 9 images for canola mapping. According to literatures, some canola indices have been proposed in relation to the spectral features of the yellow canola flower, such as the canola ratio index (CRI), normalised difference yellowness index (NDYI), canola index (CI), canola flower index (CFI), normalised difference rapeseed flower index (NRFI) and winter rapeseed index (WRI). However, it is unclear which index has more potential to extract canola based on the Landsat 9 images. Therefore, the potentials of these indices for canola mapping were quantitatively compared by using different classification methods including supervised and unsupervised classification methods. For supervised classification methods and the CFI, the overall accuracy of canola mapping was above 90% and the kappa coefficient was above 0.8. For unsupervised classification methods, CFI also performs best. It was demonstrated that the CFI outperformed the other five indices for canola mapping. It was also confirmed that Landsat 9 images and the CFI exhibit promising potential for canola mapping via quantitative comparison.

Glaciers are highly responsive to climate variations, yet monitoring them in the rugged Himalaya region poses significant challenges. This study explores the effectiveness and cost-efficiency of using machine learning models integrated with remote sensing data from Google Earth Engine (GEE) to map glacier accumulation (snow) zones in the Pārbati Valley. We tested various machine learning algorithms, including Otsu (image segmentation), K-means and cascade K-means (unsupervised classification), and random forest, minimum distance, smile CART, naive Bayes, robust tree, and support vector machine (supervised classification). Our analysis shows that the Otsu method, along with K-means, cascade K-means, and all supervised classification methods except smile CART and naive Bayes, perform similarly in mapping snowlines. Notably, the Otsu method achieved a maximum predictable error of 57 meters, which is a substantial improvement over traditional methods and indicates higher accuracy in snowline mapping. The study reveals that the regional snowline in the Pārbati Valley ranged between 5048 meters and 5113 meters during the study period. Given its superior performance, the Otsu method is recommended for identifying snowline altitudes across a wide range of glaciers in the Himalayas. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicles equipped with hyperspectral imagers have emerged as an essential technology for the characterization of inland water bodies. The high spectral and spatial resolutions of these systems enable the retrieval of a plethora of optically active water quality parameters via band ratio algorithms and machine learning methods. However, fitting and validating these models requires access to sufficient quantities of in situ reference data which are time-consuming and expensive to obtain. In this study, we demonstrate how Generative Topographic Mapping (GTM), a probabilistic realization of the self-organizing map, can be used to visualize high-dimensional hyperspectral imagery and extract spectral signatures corresponding to unique endmembers present in the water. Using data collected across a North Texas pond, we first apply GTM to visualize the distribution of captured reflectance spectra, revealing the small-scale spatial variability of the water composition. Next, we demonstrate how the nodes of the fitted GTM can be interpreted as unique spectral endmembers. Using extracted endmembers together with the normalized spectral similarity score, we are able to efficiently map the abundance of nearshore algae, as well as the evolution of a rhodamine tracer dye used to simulate water contamination by a localized source. [ABSTRACT FROM AUTHOR]

Over the last 150–200 y, urbanization and agricultural development have contributed to the loss of >99% of coastal prairie that once spanned >1 million ha of land in Louisiana. Given the extent of loss, fragmented nature of extant coastal prairie, and current threats (e.g., incompatible grazing practices, fire suppression, human disturbance, invasive species), identifying locations of unknown coastal prairie is necessary to preserve this critically imperiled ecosystem. We used remotely sensed data to identify potential locations of coastal prairie in southwestern Louisiana, USA. Given similarities between coastal prairie and other land cover types (e.g., pasture) in the region, the small number of locations available for use as training data, and the likelihood that any previously undiscovered remnants are quite small, we created two separate unsupervised classification models for our study area—one based on a Normalized Difference Vegetation Index that we calculated from 2019 NAIP imagery and one based on an Enhanced Vegetation Index that we calculated from 2019 Sentinel-2 imagery. We examined both models separately and overlapped models to look at areas of agreement, or congruence, in the potential locations of coastal prairie. The total area of model congruence was 3733 ha within our ∼330,000 ha study area, with 53% of model congruence within Calcasieu Parish and 38% in Cameron Parish. In addition, we primarily found concentrations of model congruence in north-central Cameron Parish. The methods we outline here could help inform locations of future surveys for coastal prairie, which is critical for protection and restoration of this unique ecosystem. [ABSTRACT FROM AUTHOR]

Seagrass ecosystems are pivotal in marine environments, serving as crucial habitats for diverse marine species and contributing significantly to carbon sequestration. Accurate classification of seagrass species from underwater images is imperative for monitoring and preserving these ecosystems. This paper introduces Unsupervised Curriculum Learning (UCL) to seagrass classification using the DeepSeagrass dataset. UCL progressively learns from simpler to more complex examples, enhancing the model's ability to discern seagrass features in a curriculum-driven manner. Experiments employing state-of-the-art deep learning architectures, convolutional neural networks (CNNs), show that UCL achieved overall 90.12 % precision and 89 % recall, which significantly improves classification accuracy and robustness, outperforming some traditional supervised learning approaches like SimCLR, and unsupervised approaches like Zero-shot CLIP. The methodology of UCL involves four main steps: high-dimensional feature extraction, pseudo-label generation through clustering, reliable sample selection, and fine-tuning the model. The iterative UCL framework refines CNN's learning of underwater images, demonstrating superior accuracy, generalization, and adaptability to unseen seagrass and background samples of undersea images. The findings presented in this paper contribute to the advancement of seagrass classification techniques, providing valuable insights into the conservation and management of marine ecosystems. The code and dataset are made publicly available and can be assessed here: https://github.com/nabid69/Unsupervised-Curriculum-Learning—UCL.

The subject of this article is land cover classification based on geospatial data. The supervised classification methods are appropriate for most of the thematic tasks of remote sensing because they provide the opportunity to set the characteristics of the initial classes in the form of a training sample set, in contrast to unsupervised methods. There are many approaches to processing such a set; however, their common disadvantage is that they do not consider the factor of training sample separability. This characteristic indicates the extent to which signatures representing different classes do not overlap. A low degree of separability is inherent in high-level training sample mixing. Thus, separability affects classification accuracy. One possible ways to increase separability is training sample clustering. Considering the above, the goal of this study is to develop a training sample clustering technique to improve land cover classification accuracy by increasing the separability of training samples. The tasks of this work are as follows: 1) develop a method for training sample separability assessment; 2) develop a training sample clustering technique based on training sample separability; 3) test the effectiveness of the developed technique by applying it to experimental land cover classification. In the experiments, two land cover classifications were obtained for each of the two selected study areas (i.e., one before and another after training sample clustering. Six land cover classes were defined for each experiment. The training samples were selected for each class. Conclusions. After the application of the developed technique, an increase in the separability of the training samples was evidenced by the developed separability index. In turn, this approach led to an improvement in land cover classification. For the first experiment, this was evidenced by an increase in the overall accuracy and kappa coefficient by 20% (from 63 to 83%) and 21% (from 60% to 81%), respectively. In the second experiment, the increase was 4% (from 77% to 81%) and 5% (from 66% to 71%), respectively.

Spike sorting, an indispensable process in the analysis of neural biosignals, aims to segregate individual action potentials from mixed recordings. This study delves into a comprehensive investigation of diverse unsupervised classification algorithms, some of which, to the best of our knowledge, have not previously been used for spike sorting. The methods encompass Principal Component Analysis (PCA), K-means, Self-Organizing Maps (SOMs), and hierarchical clustering. The research draws insights from both macaque monkey and human pancreatic signals, providing a holistic evaluation across species. Our research has focused on the utilization of the aforementioned methods for the sorting of 327 detected spikes within an in vivo signal of a macaque monkey, as well as 386 detected spikes within an in vitro signal of a human pancreas. This classification process was carried out by extracting statistical features from these spikes. We initiated our analysis with K-means, employing both unmodified and normalized versions of the features. To enhance the performance of this algorithm, we also employed Principal Component Analysis (PCA) to reduce the dimensionality of the data, thereby leading to more distinct groupings as identified by the K-means algorithm. Furthermore, two additional techniques, namely hierarchical clustering and Self-Organizing Maps, have also undergone exploration and have demonstrated favorable outcomes for both signal types. Across all scenarios, a consistent observation emerged: the identification of six distinctive groups of spikes, each characterized by distinct shapes, within both signal sets. In this regard, we meticulously present and thoroughly analyze the experimental outcomes yielded by each of the employed algorithms. This comprehensive presentation and discussion encapsulate the nuances, patterns, and insights uncovered by these algorithms across our data. By delving into the specifics of these results, we aim to provide a nuanced understanding of the efficacy and performance of each algorithm in the context of spike sorting. [ABSTRACT FROM AUTHOR]

This study examines the application of unsupervised classification techniques in the mapping of forest vegetation, aiming to align vegetation cover with the Braun-Blanquet classification system through remote sensing. By leveraging Landsat 8 and 9 satellite imagery and advanced clustering algorithms, specifically the Weka X-Means, this research addresses the challenge of minimizing researcher subjectivity in vegetation mapping. The methodology incorporates a two-step clustering approach to accurately classify forest communities, utilizing a comprehensive set of vegetation indices to distinguish between different types of forest ecosystems. The validation of the classification model relied on a detailed analysis of over 17,000 relevés from the "Flora" database, ensuring a high degree of accuracy in matching satellite-derived vegetation classes with field observations. The study's findings reveal the successful identification of 44 forest community types that was aggregated into seven classes of Braun-Blanquet classification system, demonstrating the efficacy of unsupervised classification in generating reliable vegetation maps. This work not only contributes to the advancement of remote sensing applications in ecological research, but also provides a valuable tool for natural resource management and conservation planning. The integration of unsupervised classification with the Braun-Blanquet system presents a novel approach to vegetation mapping, offering insights into ecological characteristics, and can be good starter point for sequestration potential of forest communities' assessment in the Republic of Tatarstan. [ABSTRACT FROM AUTHOR]

The subject of this article is land cover classification based on geospatial data. The supervised classification methods are appropriate for most of the thematic tasks of remote sensing because they provide the opportunity to set the characteristics of the initial classes in the form of a training sample set, in contrast to unsupervised methods. There are many approaches to processing such a set; however, their common disadvantage is that they do not consider the factor of training sample separability. This characteristic ind icates the extent to which signatures representing different classes do not overlap. A low degree of separability is inherent in high -level training sample mixing. Thus, separability affects classification accuracy. One possible ways to increase separability is training sample clustering. Considering the above, the goal of this study is to develop a training sample clustering technique to improve land cover classification accuracy by increasing the separability of training samples. The tasks of this work are as follows: 1) develop a method for training sample separability assessment; 2) develop a training sample clustering technique based on training sample separability; 3) test the effectiveness of the developed technique by applying it to experimental land cover classification. In the experiments, two land cover classifications were obtained for each of the two selected study areas (i.e., one before and another after training sample clustering. Six land cover classes were defined for each experiment. The training samples were selected for each class. Conclusions. After the application of the developed technique, an increase in the separability of the training samples was evidenced by the developed separability index. In turn, this approach led to an improvement in land cover classification. For the first experiment, this was evidenced by an increase in the overall accuracy and kappa coefficient by 20% (from 63 to 83%) and 21% (from 60% to 81%), respectively. In the second experiment, the increase was 4% (from 77% to 81%) and 5% (from 66% to 71%), respectively. [ABSTRACT FROM AUTHOR]

DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is an unsupervised clustering algorithm designed to identify clusters of various shapes and sizes in noisy datasets by pinpointing core points. The primary challenges associated with the DBSCAN algorithm involve the recognition of meaningful clusters within varying densities datasets and its sensitivity to parameter values of Epsilon distance and minimum number of neighbor points. These two issues may result in merging small clusters into larger clusters or splitting valid clusters into smaller clusters. A new Semi-Supervised DBSCAN (SS-DBSCAN) algorithm is introduced to improve the recognition of meaningful clusters. DBSCAN requires core points to be within, at most, Epsilon distance from their minimum neighboring points. The SS-DBSCAN algorithm, a modified version of the original DBSCAN, adds a pre-specified condition or constraint to identify core points further. This extra constraint is related to the clustering objective of a given dataset. To evaluate the effectiveness of SS-DBSCAN, we utilize three datasets: letter recognition, wireless localization, and Modern Standard Arabic (MSA) combined with Iraqi words language modeling. V-measure is used to evaluate the clustering efficiency for the letters recognition and wireless localization datasets. The perplexity (pp) of the class-based language model, built on the produced clusters, is the metric used for the Iraqi-MSA dataset clustering effectiveness. Experimental results showed the significant effectiveness of SS-DBSCAN. It outperforms DBSCAN when applied to letters and Iraqi-MSA datasets with improvements of 65% and 14.5%, respectively. A comparable performance was achieved when clustering the wireless localization dataset. Additionally, to assess the effectiveness of SS-DBSCAN, its performance has been compared to various modified versions of DBSCAN using four metrics: V-measure, PP, Adjusted Rand Index (ARI), and the Silhouette score. Based on these metrics, the results showed that SS-DBSCAN outperformed most DBSCAN versions in three case studies. Consequently, the proposed SS-DBSCAN algorithm is particularly suitable for high-density datasets. The SS-DBSCAN python code is available at https://github.com/TibaZaki/SS_DBSCAN.

Polarimetric synthetic aperture radar (PolSAR) image classification is a critical application of remote sensing image interpretation. Most of the early algorithms that use hand-crafted features to divide the image into various scattering categories have a general classification performance. The convolutional neural networks (CNNs) show superior performance in image processing with powerful nonlinear representation capabilities. However, they also require a large amount of labeled training data, which limit the practical use of CNNs in PolSAR image classification scene where labeled data are rare and expensive. To address the previous issue, this article proposes a deep similarity clustering model with compound regularization for unsupervised PolSAR image classification. The proposed model combines an unsupervised feature extraction pipeline with Wishart distance metric and a deep clustering pipeline with feature similarity metric. The regularization combines two ingredients to preserve both the sharpness of edges and the semantic continuity of the image contents. The first is the differential constraint based on pixel-level features, and the second is the graph partition constraint based on superpixel-level features. Experiments prove the effectiveness of the proposed method on both spaceborne (RadarSat2) and airborne (E-SAR/UAVSAR) PolSAR images. The visual results and quantitative scores show that our method outperforms the traditional unsupervised methods and deep-learning-based unsupervised methods with a large margin.

Building detection plays an important role in urban applications and is usually a prerequisite for contour extraction and building modeling. Over the last decades, airborne LiDAR data have been used due to its capability to represent terrestrial surfaces and objects with high geometric quality. In this paper, it is proposed a novel building detection approach based on geometric/morphological object characteristics. The proposed strategy is divided into three main stages: 1) selection of candidate points based on height; 2) building detection using the geometric feature (omnivariance) and K-means clustering algorithm; and 3) refinement based on majority filter and mathematical morphology. The experiments were conducted using airborne LiDAR datasets with varying point density acquired in different urban environments. The results indicated the robustness of the proposed approach for all datasets and environmental complexities, presenting average Fscore of around 96%. In addition, the results pointed out that point density can impact the building detection, producing better results for higher point density datasets. Compared with related approaches, the proposed strategy results in better performance in terms of completeness, producing an omission error rate smaller than 3%.

Supervised polarimetric synthetic aperture radar (PolSAR) image classification demands a large amount of precisely labeled data. However, such data are difficult to obtain. Therefore, many unsupervised methods have been proposed for unsupervised PolSAR image classification. The classification maps of unsupervised methods contain many high-confidence samples. These samples, which are often ignored, can be used as supervisory information to improve classification performance on PolSAR images. This study proposes a new unsupervised PolSAR image classification framework. The framework combines high-confidence superpixel pseudo-labeled samples and semi-supervised classification methods. The experiments indicated that this framework could achieve higher-level effectiveness in unsupervised PolSAR image classification. First, superpixel segmentation was performed on PolSAR images, and the geometric centers of the superpixels were generated. Second, the classification maps of rotation-domain deep mutual information (RDDMI), an unsupervised PolSAR image classification method, were used as the pseudo-labels of the central points of the superpixels. Finally, the unlabeled samples and the high-confidence pseudo-labeled samples were used to train an excellent semi-supervised method, similarity matching (SimMatch). Experiments on three real PolSAR datasets illustrated that, compared with the excellent RDDMI, the accuracy of the proposed method was increased by 1.70%, 0.99%, and 0.8%. The proposed framework provides significant performance improvements and is an efficient method for improving unsupervised PolSAR image classification.

A family of parsimonious contaminated shifted asymmetric Laplace mixtures is developed for unsupervised classification of asymmetric clusters in the presence of outliers and noise. A series of constraints are applied to a modified factor analyzer structure of the component scale matrices, yielding a family of twelve models. Application of the modified factor analyzer structure and these parsimonious constraints makes these models effective for the analysis of high-dimensional data by reducing the number of free parameters that need to be estimated. A variant of the expectation-maximization algorithm is developed for parameter estimation with convergence issues being discussed and addressed. Popular model selection criteria like the Bayesian information criterion and the integrated complete likelihood (ICL) are utilized, and a novel modification to the ICL is also considered. Through a series of simulation studies and real data analyses, that includes comparisons to well-established methods, we demonstrate the improvements in classification performance found using the proposed family of models. [ABSTRACT FROM AUTHOR]

Natural disasters, especially those related to water—like storms and floods—have increased over the last decades both in number and intensity. Under the current Climate Change framework, several reports predict an increase in the intensity and duration of these extreme climatic events, where the Mediterranean area would be one of the most affected. This paper develops a decision support system based on Bayesian inference able to predict a flood alert in Andalusian Mediterranean catchments. The key point is that, using simple weather forecasts and live measurements of river level, we can get a flood-alert several hours before it happens. A set of models based on Bayesian networks was learnt for each of the catchments included in the study area, and joined together into a more complex model based on a rule system. This final meta-model was validated using data from both non-extreme and extreme storm events. Results show that the methodology proposed provides an accurate forecast of the flood situation of the greatest catchment areas of Andalusia. [ABSTRACT FROM AUTHOR]

Nowadays, an enormous amount of high dimensional data is available in the field of neuroscience. Handling these data is complex and requires the use of efficient tools to transform them into useful knowledge. In this work we present NeuroSuites, an easy-access web platform with its own architecture. We compare our platform with other software currently available, highlighting its main strengths. Thanks to its defined architecture, it is able to handle large-scale problems common in some neuroscience fields. NeuroSuites has different neuroscience-oriented applications and tools to integrate statistical data analysis and machine learning algorithms commonly used in this field. As future work, we want to further expand the list of available software tools as well as improve the platform interface according to user demands. [ABSTRACT FROM AUTHOR]

Vocal individuality has been widely documented in the Tawny owl (Strix aluco); however, all statistical tools employed thus far to discriminate individual vocalisations have relied on prior knowledge regarding number and identity of individuals. In this study, we tested the effectiveness of four unsupervised clustering algorithms in distinguishing among eight Tawny owl males, solely based on acoustic characteristics of their vocalisations. We also employed both traditional bound-based and robust measurements of acoustic signal to compare their efficacy. We finally evaluated the applicability of this method in identifying the number and distribution of the remaining males recorded in our study area. Three of the four unsupervised techniques had a high rate of success in discriminating among vocalisations of the eight males. In all cases, the best results were obtained using robust measurements. However, when extending the analysis to the remaining unknown males recorded, the highest rate of misclassification errors made results more difficult to interpret. Our study provided a useful tool to discriminate male Tawny owls when only their call recordings are available. Furthermore, this method could be extended to other nocturnal and vociferous species, representing one of the few existing approaches for unsupervised classification of individuals based on acoustic features. [ABSTRACT FROM AUTHOR]

The data science discipline can play a crucial role in developing effective data driven strategies for the valorization and promotion of the cultural heritage (CH) domain. Machine learning approaches can provide new perspectives, allowing knowledge extraction and insights generation from data since in the last decade CH domain has benefited from the applications of internet of things (IoT) solutions in order to improve visitors' experience. Analyzing a great amount of data increasingly requires the use of advanced mathematical algorithms and therefore requires distribution, calculation and digital protection services. Data represent a great challenge for the CH domain, as well as a resource; this paper presents and discusses the application of a machine learning approach on IoT cultural data collected in the National Archaeological Museum of Naples. With the deployment of some Bluetooth sensing boards we collected the visit paths of the users in a non-invasive way. The research goal is to analyze and classify the collected visiting behavioural data in order to produce useful insights for cultural stakeholders. The knowledge of people behaviours can help museum organizations both in terms of medium-long term strategy and also in terms of strictly operational decisions. [ABSTRACT FROM AUTHOR]

Electrodermal Activity (EDA) has shown great potential for emotion recognition and the early detection of physiological anomalies associated with stress. However, its non-stationary nature limits the capability of current analytical and detection techniques, which are highly dependent on signal stability and controlled environmental conditions. This paper proposes a framework for EDA normalization based on the exponential moving average (EMA) with outlier removal applicable to non-stationary heteroscedastic signals and a novel set of features for analysis. The normalized time series preserves the morphological and statistical properties after transformation. Meanwhile, the proposed features expand on typical time-domain EDA features and profit from the resulting normalized signal properties. Parameter selection and validation were performed using two different EDA databases on stress assessment, accomplishing trend preservation using windows between 5 and 20 s. The proposed normalization and feature extraction framework for EDA analysis showed promising results for the identification of noisy, relaxed and arousal-like patterns in data with conventional clustering approaches like K-means over the aforementioned normalized features. [ABSTRACT FROM AUTHOR]

As the volume of satellite images increases rapidly, unsupervised classification can be utilized to swiftly investigate land cover distributions without prior knowledge and to generate training data for supervised (or deep learning-based) classification. In this study, an inter-image k-means clustering algorithm (IIkMC), as an improvement of the native k-means clustering algorithm (kMC), was introduced to obtain a single set of class signatures so that the classification results could be compatible among multiple images. Because IIkMC was a computationally intensive algorithm, parallelized approaches were deployed, using multi-cores of a central processing unit (CPU) and a graphics processing unit (GPU), to speed up the process. kMC and IIkMC were applied to a series of images acquired in a PlanetScope mission. In addition to the capability of the inter-image compatibility of the classification results, IIkMC could settle the problem of incomplete segmentation and class canceling revealed in kMC. Based on CPU parallelism, the speed of IIkMC improved, becoming up to 12.83 times better than sequential processing. When using a GPU, the speed improved up to 25.53 times, rising to 39.00 times with parallel reduction. From the results, it was confirmed IIkMC provided more reliable results than kMC, and its parallelism could facilitate the overall inspection of multiple images. [ABSTRACT FROM AUTHOR]

Parkinson, F.; Douglas, K.; Li, Z.; Meijer, A.; Stacey, C.D.; Kung, R., and Podhorodeski, A., 2024. A generalized semiautomated method for seabed geology classification using multibeam data and maximum likelihood classification. Journal of Coastal Research, 40(1), 1–16. Charlotte (North Carolina), ISSN 0749-0208. This paper presents a GIS-based model to perform semiautomated seabed classification that can act as a first-pass, pseudoclassified surficial geological map. The user can then edit the output into a finalized map in less time than by manual classification. The model uses maximum likelihood classification with unsupervised classification through iterative self-organizing clusters. This model is fully contained within the ArcGIS software suite as a ModelBuilder workflow composed of geoprocessing tools and Python script tools. Model inputs tested include different combinations of multibeam echosounder–derived data: slope, backscatter, and terrain ruggedness. Furthermore, to test the assumption of Gaussian distribution of input data required for maximum likelihood classification, Box–Cox power transformations were applied to slope and backscatter data and were used as model inputs. To illustrate the performance of the model, two locations are highlighted as case studies: Milbanke Sound and Spiller Channel, located on the central coast of British Columbia, Canada. Association between model outputs and ground-truth classes was generally weak to moderate when measured using Cramér's V association scores. Overall, the slope and backscatter parameter model had the highest scores of association. Results from an overlay analysis comparing model outputs with user-confirmed polygons show that the slope and backscatter model performs best in regions with distinct changes in the hardness of sediments but that in fjord regions dominated geologically by steeper bathymetric change, the slope parameter model may perform better. However, all model outputs had difficulty delineating bedrock units. The model has the flexibility to identify certain seabed habitat features as well, including glass sponge reefs—biologically active bioherms that have led to marine protected area designations in other areas of British Columbia. [ABSTRACT FROM AUTHOR]

The proposed facies classification approach uses three features of the GLCM with multitemporal X-band SAR data of TerraSAR-X combined with k-means clustering. This approach was tested on 15 transboundary Himalayan glaciers. They were classified into six facies namely debris, dry snow, wet snow, bare ice, percolation zone 1 and percolation zone 2 and the covered areas for these classes are 20.56%, 12.50%, 20.26%, 13.16%, 25.49% and 8.03% of total area respectively. After classification, the signature for classes are validated using the benchmark outline of East Rathong glacier, India. The accuracy achieved is 89.56% with negligible variance and kappa of 0.87. [ABSTRACT FROM AUTHOR]

Background: The cloud-penetrating and fog-penetrating capability of Synthetic Aperture Radar (SAR) give it the potential for application in forest fire progress monitoring; however, the low extraction accuracy and significant salt-and-pepper noise in SAR remote sensing mapping of the burned area are problems. Aims: This paper provides a method for accurately extracting the burned area based on fully exploiting the changes in multiple different dimensional feature parameters of dual-polarised SAR images before and after a fire. Methods: This paper describes forest fire progress monitoring using dual-polarisation SAR images combined with multi-scale segmentation and unsupervised classification. We first constructed polarisation feature and texture feature datasets using multi-scene Sentinel-1 images. A multi-scale segmentation algorithm was then used to generate objects to suppress the salt-and-pepper noise, followed by an unsupervised classification method to extract the burned area. Key results: The accuracy of burned area extraction in this paper is 91.67%, an improvement of 33.70% compared to the pixel-based classification results. Conclusions: Compared with the pixel-based method, our method effectively suppresses the salt-and-pepper noise and improves the SAR burned area extraction accuracy. Implications: The fire monitoring method using SAR images provides a reference for extracting the burned area under continuous cloud or smoke cover. This paper describes a method to monitor forest fire progress using dual-polarisation Synthetic Aperture Radar (SAR) images combined with multi-scale segmentation and unsupervised classification. We aimed to take full advantage of the many different dimensions of feature parameter changes caused by forest fires, relying on time-series dual-polarised SAR imagery to achieve burned area extraction and forest fire progress monitoring. [ABSTRACT FROM AUTHOR]

Erosion-induced lateral soil redistribution leads to spatially heterogenous soil composition, which can be captured through the distinctive spectral reflectance of soils under varying levels of erosion influence. This points to the potential of using remotely sensed soil spectra to detect severe erosion and deposition hotspots in exposed croplands and, importantly, further differentiate the intra-class spectral variability of moderate erosion that often occupies the largest proportion. Here, we aim to develop a two-step erosion classification and mapping approach based on multitemporal compositing of Sentinel-2 bare soil images of a typical agricultural region (11,500 km2) of northeast China. A random forest classifier was firstly trained against the ground-truth data derived from very high resolution (VHR) imagery in Google Earth, with an overall accuracy of 91 % that allowed for clear delineation of severe erosion and deposition areas based on their distinct topographic and spectral features particularly in the red and red-edge bands. In the second step, the remaining area of moderate erosion (60.30 %) was further differentiated using Iterative Self-Organizing cluster unsupervised classification to yield a five-class soil erosion map at 10 m spatial resolution. The accuracy of the predicted map was successfully validated by independent Caesium-137 (137Cs) and soil organic carbon observations at catchment and regional scales, as revealed by significant inter-class differences in soil redistribution rates estimated from 137Cs inventory. The severe erosion class had a soil loss rate of 5.5 mm yr−1, suggesting that previous assessments have underestimated erosion severity. The spatial accordance of crop growth with soil erosion intensity, particularly in localized settings, further highlighted the potential of bare soil imaging for mapping the spatiotemporal development of soil erosion and its response to targeted sustainable cropland management efforts.

Unmanned aerial vehicles equipped with hyperspectral imagers have emerged as an essential technology for the characterization of inland water bodies. The high spectral and spatial resolutions of these systems enable the retrieval of a plethora of optically active water quality parameters via band ratio algorithms and machine learning methods. However, fitting and validating these models requires access to sufficient quantities of in situ reference data which are time-consuming and expensive to obtain. In this study, we demonstrate how Generative Topographic Mapping (GTM), a probabilistic realization of the self-organizing map, can be used to visualize high-dimensional hyperspectral imagery and extract spectral signatures corresponding to unique endmembers present in the water. Using data collected across a North Texas pond, we first apply GTM to visualize the distribution of captured reflectance spectra, revealing the small-scale spatial variability of the water composition. Next, we demonstrate how the nodes of the fitted GTM can be interpreted as unique spectral endmembers. Using extracted endmembers together with the normalized spectral similarity score, we are able to efficiently map the abundance of nearshore algae, as well as the evolution of a rhodamine tracer dye used to simulate water contamination by a localized source.

The flagellar movement of the mammalian sperm plays a crucial role in fertilization. In the female reproductive tract, human spermatozoa undergo a process called capacitation which promotes changes in their motility. Only capacitated spermatozoa may be hyperactivated and only those that transition to hyperactivated motility are capable of fertilizing the egg. Hyperactivated motility is characterized by asymmetric flagellar bends of greater amplitude and lower frequency. Historically, clinical fertilization studies have used two-dimensional analysis to classify sperm motility, despite the inherently three-dimensional (3D) nature of sperm motion. Recent research has described several 3D beating features of sperm flagella. However, the 3D motility pattern of hyperactivated spermatozoa has not yet been characterized. One of the main challenges in classifying these patterns in 3D is the lack of a ground-truth reference, as it can be difficult to visually assess differences in flagellar beat patterns. Additionally, it is worth noting that only a relatively small proportion, approximately 10-20% of sperm incubated under capacitating conditions exhibit hyperactivated motility. In this work, we used a multifocal image acquisition system that can acquire, segment, and track sperm flagella in 3D+t. We developed a feature-based vector that describes the spatio-temporal flagellar sperm motility patterns by an envelope of ellipses. The classification results obtained using our 3D feature-based descriptors can serve as potential label for future work involving deep neural networks. By using the classification results as labels, it will be possible to train a deep neural network to automatically classify spermatozoa based on their 3D flagellar beating patterns. We demonstrated the effectiveness of the descriptors by applying them to a dataset of human sperm cells and showing that they can accurately differentiate between non-hyperactivated and hyperactivated 3D motility patterns of the sperm cells. This work contributes to the understanding of 3D flagellar hyperactive motility patterns and provides a framework for research in the fields of human and animal fertility.

Colored dissolved organic matter (CDOM) is a significant constituent of aquatic systems and biogeochemical cycles. Satellite CDOM retrievals are challenging in inland waters, due to overlapped absorption properties of bio-optical parameters, like Total Suspended Matter (TSM). In this framework, we defined an accurate CDOM model using Sentinel2-MSI (S2-MSI) data in Pertusillo Lake (Southern Italy) adopting a classification scheme based on satellite TSM data. Empirical relationships were established between the CDOM absorption coefficient, aCDOM (440), and reflectance band ratios using ground-based measurements. The Green-to-Red (B3/B4 and B3/B5) and Red-to-Blue (B4/B2 and B5/B2) band ratios showed good relationships (R2 ≥ 0.75), which were further improved according to sub-region division (R2 up to 0.93). The best accuracy of B3/B4 in the match-ups between S2-MSI-derived and in situ band ratios proved the exportability on S2-MSI data of two B3/B4-based aCDOM (440) models, namely the fixed (for the whole PL) and the switching one (according to sub-region division). Although they both exhibited good agreements in aCDOM (440) retrievals (R2 ≥ 0.69), the switching model showed the highest accuracy (RMSE of 0.0155 m−1). Finally, the identification of areas exposed to different TSM patterns can assist with refining the calibration/validation procedures to achieve more accurate aCDOM (440) retrievals. [ABSTRACT FROM AUTHOR]

Data are needed to assess the relationships between urbanization and biodiversity to establish conservation priorities. However, many of these relationships are difficult to fully assess using traditional research methods. To address this gap and evaluate new acoustic sensors and associated data, we conducted a multimethod analysis of biodiversity in a rapidly urbanizing county: Greenville, South Carolina, USA. We conducted audio recordings at 25 points along a development gradient. At the same locations, we used refugia tubes, visual assessments, and an online database. Analysis focused on species identification of both audio and visual data at each point along the trail to determine relationships between both herpetofauna and acoustic indices (as proxies for biodiversity) and environmental gradient of land use and land cover. Our analysis suggests the use of a multitude of different sampling methods to be conducive to the completion of a more comprehensive occupancy measure. Moving forward, this research protocol can potentially be useful in the establishment of more effective wildlife occupancy indices using acoustic sensors to move toward future conservation policies and efforts concerning urbanization, forest fragmentation, and biodiversity in natural, particularly forested, ecosystems. [ABSTRACT FROM AUTHOR]

Accurate and up-to-date land use information is essential for effective urban planning and environmental management. This paper presents a methodology for the unsupervised classification of Casablanca's land cover using Google Earth Engine (GEE). The study exploits multispectral satellite imagery, in particular Landsat data, to extract meaningful land-use classes without the need for manual labeling. The workflow includes data collection, pre-processing, unsupervised clustering, and visualization of results. By applying the k-means and Learning Vector Quantization (LVQ) clustering algorithms, the city's land area is divided into distinct clusters, each representing a specific land-use class. The resulting land-use map provides valuable information on Casablanca's urban landscape, highlighting forest areas, crops, built-up infrastructure, water bodies, and barren areas. This automated approach demonstrates the potential of GEE as a powerful tool for land use analysis, enabling effective, data-driven decision-making for urban development and environmental monitoring. The methodology presented can serve as a basis for similar studies in other regions, contributing to the advancement of remote sensing and geospatial analysis techniques for urban and environmental studies. This study evaluates the effectiveness of these two algorithms in terms of overall accuracy and kappa coefficient. The K-means algorithm recorded moderate accuracy. The LVQ algorithm, on the other hand, performed the least well. [ABSTRACT FROM AUTHOR]

Unsupervised classification is used in credit risk assessment to reduce human resource costs and make informed decisions in the shortest possible time. Although several studies show that support vector machine-based methods have better performance in unlabeled datasets, several factors still negatively affect these models, such as unstable results due to random initialization, reduced effectiveness due to kernel dependencies, and noise points and outliers. This paper introduces an unsupervised classification method based on a fuzzy unsupervised quadratic surface support vector machine without a kernel to avoid selecting related kernel parameters for credit risk assessment. In addition, we propose an innovative fuzzy membership function for reducing noise points and outliers in line with the direction of sample density variation. Fuzzy Unsupervised QSSVM (FUS-QSSVM) outperforms well-known SVM-based methods based on numerical tests on public benchmark credit data. In some real-world applications, the proposed method has significant potential as well as being effective, efficient, and robust. The algorithm can therefore increase the number of potential customers of financial institutions as well as increase profitability. [ABSTRACT FROM AUTHOR]

This article describes a comparison of the pixel-based classification methods used to distinguish ice from other land cover types. The article focuses on processing RGB imagery, as these are very easy to obtained. The imagery was taken using UAVs and has a very high spatial resolution. Classical classification methods (ISODATA and Maximum Likelihood) and more modern approaches (support vector machines, random forests, deep learning) have been compared for image data classifications. Input datasets were created from two distinct areas: The Pond Skříň and the Baroch Nature Reserve. The images were classified into two classes: ice and all other land cover types. The accuracy of each classification was verified using a Cohen's Kappa coefficient, with reference values obtained via manual surface identification. Deep learning and Maximum Likelihood were the best classifiers, with a classification accuracy of over 92% in the first area of interest. On average, the support vector machine was the best classifier for both areas of interest. A comparison of the selected methods, which were applied to highly detailed RGB images obtained with UAVs, demonstrates the potential of their utilization compared to imagery obtained using satellites or aerial technologies for remote sensing. [ABSTRACT FROM AUTHOR]