ISODATA Clustering Research Articles

Harnessing Spectral Libraries From AVIRIS-NG Data for Precise PFT Classification: A Deep Learning Approach.

The generation of spectral libraries using hyperspectral data allows for the capture of detailed spectral signatures, uncovering subtle variations in plant physiology, biochemistry, and growth stages, marking a significant advancement over traditional land cover classification methods. These spectral libraries enable improved forest classification accuracy and more precise differentiation of plant species and plant functional types (PFTs), thereby establishing hyperspectral sensing as a critical tool for PFT classification. This study aims to advance the classification and monitoring of PFTs in Shoolpaneshwar wildlife sanctuary, Gujarat, India using Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) and machine learning techniques. A comprehensive spectral library was developed, encompassing data from 130 plant species, with a focus on their spectral features to support precise PFT classification. The spectral data were collected using AVIRIS-NG hyperspectral imaging and ASD Handheld Spectroradiometer, capturing a wide range of wavelengths (400-1600 nm) to encompass the key physiological and biochemical traits of the plants. Plant species were grouped into five distinct PFTs using Fuzzy C-means clustering. Key spectral features, including band reflectance, vegetation indices, and derivative/continuum properties, were identified through a combination of ISODATA clustering and Jeffries-Matusita (JM) distance analysis, enabling effective feature selection for classification. To assess the utility of the spectral library, three advanced machine learning classifiers-Parzen Window (PW), Gradient Boosted Machine (GBM), and Stochastic Gradient Descent (SGD)-were rigorously evaluated. The GBM classifier achieved the highest accuracy, with an overall accuracy (OAA) of 0.94 and a Kappa coefficient of 0.93 across five PFTs.

Data-driven distributionally robust optimization approach for the coordinated dispatching of the power system considering the correlation of wind power

With the increasing penetration of large-scale wind power into the power grid, it is crucial to develop a precise model to accurately depict the stochasticity and correlation among wind farm outputs, which is highly important for ensuring the safe and efficient utilization of wind energy in grid dispatching. In this article, a data-driven distributionally robust optimization (DDRO) dispatching approach that accounts for spatial correlations among outputs from multiple wind farms is proposed. The proposed approach is applied to a source-network-load-storage grid system to ascertain unit start-stop schedules and resource allocation effectively. First, a truncated spatial correlation model is proposed, enabling a comprehensive representation of spatial correlations and output constraints between distinct wind farms. Second, the ISODATA clustering algorithm is employed to generate typical scenarios, reduce model complexity, and expedite the computation process. Third, a unit commitment model considering the demand response is constructed and solved using the DDRO approach. Finally, the proposed model is applied to the IEEE 30-bus system to test its robustness and cost-effectiveness compared to the traditional robust optimization model. Additionally, it is applied to the IEEE 118-bus system to demonstrate its scalability and stability.

Seagrass ecosystem biodiversity mapping in part of Rote Island using multi-generation PlanetScope imagery

Remote sensing offers an effective and efficient solution to provide information on the biodiversity of seagrass ecosystems, which is currently lacking in most parts of the world. Therefore, this study aimed to map the biodiversity of seagrass ecosystems in parts of Rote Island, which is one of the seagrass biodiversity hotspots, using multi-generation PlanetScope magery to see how they compare. The most frequently used biodiversity indicators were identified, including the major benthic habitat (coral, seagrass, macroalgae, bare substrate) and the composition of seagrass species based on life forms. We also aim to understand the actual biodiversity indicators of seagrass ecosystems captured by PlanetScope imagery. To achieve this, field data was integrated with the resulting ISODATA classification results to assess what ISODATA class clusters represent in the field, and new classification schemes are developed accordingly. The random forest algorithm was used to carry out the classification, with seagrass field data serving as training data. Independent field data was subsequently used to assess the accuracy. The results showed that the accuracy of benthic habitat and seagrass mapping ranged from 60%-70%. However, through the use of a classification scheme built on ISODATA clustering, the spatial distribution of classes and accuracy of all PlanetScope images was significantly improved to > 90%. This highlighted the importance of understanding which indicators of seagrass biodiversity were effectively captured by PlanetScope images to achieve higher mapping accuracy. Overall, this approach optimized the ability of PlanetScope images to map seagrass biodiversity while obtaining a higher number of biodiversity indicator classes and mapping accuracy than the commonly used biodiversity indicator classification scheme.

Seagrass ecosystem biodiversity mapping in part of Rote Island using multi-generation PlanetScope imagery

Remote sensing offers an effective and efficient solution to provide information on the biodiversity of seagrass ecosystems, which is currently lacking in most parts of the world. Therefore, this study aimed to map the biodiversity of seagrass ecosystems in parts of Rote Island, which is one of the seagrass biodiversity hotspots, using multi-generation PlanetScope magery to see how they compare. The most frequently used biodiversity indicators were identified, including the major benthic habitat (coral, seagrass, macroalgae, bare substrate) and the composition of seagrass species based on life forms. We also aim to understand the actual biodiversity indicators of seagrass ecosystems captured by PlanetScope imagery. To achieve this, field data was integrated with the resulting ISODATA classification results to assess what ISODATA class clusters represent in the field, and new classification schemes are developed accordingly. The random forest algorithm was used to carry out the classification, with seagrass field data serving as training data. Independent field data was subsequently used to assess the accuracy. The results showed that the accuracy of benthic habitat and seagrass mapping ranged from 60%-70%. However, through the use of a classification scheme built on ISODATA clustering, the spatial distribution of classes and accuracy of all PlanetScope images was significantly improved to > 90%. This highlighted the importance of understanding which indicators of seagrass biodiversity were effectively captured by PlanetScope images to achieve higher mapping accuracy. Overall, this approach optimized the ability of PlanetScope images to map seagrass biodiversity while obtaining a higher number of biodiversity indicator classes and mapping accuracy than the commonly used biodiversity indicator classification scheme.

Imbalanced Data Over-Sampling Method Based on ISODATA Clustering

Imbalanced Data Over-Sampling Method Based on ISODATA Clustering

A Novel Remote Sensing Image Enhancement Method, the Pseudo-Tasseled Cap Transformation: Taking Buildings and Roads in GF-2 as an Example

With the improvements in sensor accuracy, the spectral features of high-resolution remote sensing images become more complex. As a result, the classification accuracy for land cover classification decreases. Remote sensing image enhancements can improve the visual effect and the intra-class consistency and enhance the characteristics of ground objects. These enhancements are important for both image interpretation and improving image segmentation accuracy. In this study, we propose a pseudo-tasseled cap transformation (pseudo-TCT) through an orthogonal linear transformation of Gaofen-2 (GF-2) images using the untransposed tasseled cap transformation (TCT) coefficients, and further, enhance the visual effect and the separability among ground objects by linear stretching and percentage truncation stretching. To examine the separability among ground objects in the pseudo-TCT image, we used K-Means clustering, ISODATA clustering and 3D visualization of the spectral features of typical ground objects. The results show that the separability of buildings and roads from background objects is better than in the original image and the TCT image, and typical ground objects are effectively distinguished. Additionally, we visualized intra-class consistency by calculating the mean Euclidean distance between the pixel values of each point and the pixel values of its eight neighboring points and calculated the standard deviation of the intra-class consistency images. The results indicate that the secondary textures of the objects were weakened, and edges were made clearer, enhancing intra-class consistency. The pseudo-TCT is effective, at least in our work, and could be a candidate for image enhancement under certain applications.

Stability Analysis of Unmixing-Based Spatiotemporal Fusion Model: A Case of Land Surface Temperature Product Downscaling

The unmixing-based spatiotemporal fusion model is one of the effective ways to solve limitations in temporal and spatial resolution tradeoffs in a single satellite sensor. By using fusion data from different satellite platforms, high resolution in both temporal and spatial domains can be produced. However, due to the ill-posed characteristic of the unmixing function, the model performance may vary due to the different model setups. The key factors affecting the model stability most and how to set up the unmixing strategy for data downscaling remain unknown. In this study, we use the multisource land surface temperature as the case and focus on the three major factors to analyze the stability of the unmixing-based fusion model: (1) the definition of the homogeneous change regions (HCRs), (2) the unmixing levels, and (3) the number of HCRs. The spatiotemporal data fusion model U-STFM was used as the baseline model. The results show: (1) The clustering-based algorithm is more suitable for detecting HCRs for unmixing. Compared with the multi-resolution segmentation algorithm and k-means algorithm, the ISODATA clustering algorithm can more accurately describe LST’s temporal and spatial changes on HCRs. (2) For the U-STFM model, applying the unmixing processing at the change ratio level can significantly reduce the additive and multiplicative noise of the prediction. (3) There is a tradeoff effect between the number of HCRs and the solvability of the linear unmixing function. The larger the number of HCRs (less than the available MODIS pixels), the more stable the model is. (4) For the fusion of the daily 30 m scale LST product, compared with STARFM and ESTARFM, the modified U-STFM (iso_USTFM) achieved higher prediction accuracy and a lower error (R 2: 0.87 and RMSE:1.09 k). With the findings of this study, daily fine-scale LST products can be predicted based on the unmixing-based spatial–temporal model with lower uncertainty and stable prediction.

Hybrid learning model for analysing the Uppal earth region, in Telangana state, using multispectral Landsat-8 OLI images

Remote Sensing (RS) and Geographical Information Systems (GIS) are being widely used to carry out analysis of the Earth's surface. In this paper, a hybrid learning model is proposed for the classification and analysis of the Uppal earth region, located nearby Hyderabad in Telangana state. In the hybrid learning model, the ISODATA clustering algorithm is combined with the Normalised Vegetation Index (NDVI) and K-means learning model. In this proposal, the spectral features of the Uppal region are extracted from the satellite images and used for further analysis. The obtained accuracy of the proposed Merged-ISODATA algorithm is 74.33% and the Kappa value is 0.64. The obtained accuracy and Kappa value for existing ISODATA clustering and K-Means algorithm are 71.5% and 0.58. These values imply that the obtained results of the proposed algorithm are better than the results obtained in existing approaches.

Planet Four: A Neural Network’s search for polar spring-time fans on Mars

Dark deposits visible from orbit appear in the Martian south polar region during the springtime. These are thought to form from explosive jets of carbon dioxide gas breaking through the thawing seasonal ice cap, carrying dust and dirt which is then deposited onto the ice as dark ‘blotches’, or blown by the surface winds into streaks or ‘fans’. We investigate machine learning (ML) methods for automatically identifying these seasonal features in High Resolution Imaging Science Experiment (HiRISE) satellite imagery. We designed deep Convolutional Neural Networks (CNNs) that were trained and tested using the catalog generated by Planet Four, an online citizen science project mapping the south polar seasonal deposits. We validated the CNNs by comparing their results with those of ISODATA (Iterative Self-Organizing Data Analysis Technique) clustering and as expected, the CNNs were significantly better at predicting the results found by Planet Four, in both the area of predicted seasonal deposits and in delineating their boundaries. We found neither the CNNs or ISODATA were suited to predicting the source point and directions of seasonal fans, which is a strength of the citizen science approach. The CNNs showed good agreement with Planet Four in cross-validation metrics and detected some seasonal deposits in the HiRISE images missed in the Planet Four catalog; the total area of seasonal deposits predicted by the CNNs was 27% larger than that of the Planet Four catalog, but this aspect varied considerably on a per-image basis.

Comparative assessment of homogeneity differences in multi-temporal NDVI strata and the currently used agricultural area frames in Rwanda

This study compared two methods used for agricultural statistics generation in Rwanda. The first method is area frame sampling, which is also the currently used method in Rwandan seasonal agricultural surveys; while the second method is the application of remote sensing technique using multi-temporal Normalised Difference Vegetation Index (NDVI) classes to stratify land into homogenous agriculture land classes. The analysis of the methodological flow of Rwanda area frames and the estimated homogeneity in the resulting frames was mainly based on literature review. For the delineation of homogeneous NDVI classes, the study used 10 years data from Moderate Resolution Imaging Spectroradiometer (MODIS) sensor (2004 – 2014). The NDVI data were classified using ISODATA clustering technique, and the focus was put on agriculture-dominated classes, obtained through the intersection with 2010 national land use and land cover data. Analysis of Variance (ANOVA) and Fisher’s Least Significant Difference (LSD) statistical methods were applied to investigate significant differences between and within NDVI classes and the currently used Rwanda strata in terms of area coverage of four (4) dominant crops in Rwanda – banana, maize, cassava, and beans. The results of the analysis revealed homogeneity of 85% within NDVI classes, and 69% within the current Rwanda strata, at p = 0.05. The NDVI classes were also used to improve the Rwanda strata, and the homogeneity has increased by 5%; reaching 74% after NDVI-based reclassification.

Effects of Monocropping on Land Cover Transitions in the Wet Evergreen Agro-Ecological Zone of Ghana

The wet evergreen forest ecological zone of Ghana is undergoing notable structural changes, although most maps erroneously depict the area to be an intact forest. Several hectares of plantation agriculture such as rubber, oil palm, and cocoa are replacing the natural vegetation and habitats, thereby threatening indigenous biodiversity. This study aimed to assess the effects of tree monocrop proliferation on landscape transitions between 1986 and 2020 in the Wassa East District of Ghana. The ISODATA clustering technique was used to produce land cover category maps from Landsat images of 1986, 2002, and 2020. A post-classification change detection technique resulted in transition matrices which were used for the computation of land cover transition intensities over 34 years. The results showed that the landscape was dominated by forests in 1986. Still, the forest continuously declined by 34% (1% annually) in 2020 while all other land cover types increased in both periods with cocoa covering 23% oil palm covering 14%, and rubber covering 2% of the land area. These conversions in the land area are important, especially since the closed forests usually transition to monocrops through the intermediate process: food crop farming. This information is essential for decision making on land development and biodiversity conservation.

Remote Sensing and GIS Based Crop Acreage Estimation of the Rabi Season Growing Crop of the Middle Gujarat (India)

Accurate and precise information about crop type, crop stage, and crop acreage is essential for sustainable utilization of available water resources. The present study is concerned with the estimation of Rabi season growing crops in the Panchmahal district of Gujarat state, India. In these study generation of spectral profiles and crop acreage estimation, Sentinel-2 satellite images were classified using unsupervised classification with ISODATA clustering classification techniques. The satellite image of the study area was classified into 52 classes and overlaid with a ground truth point shape file on the classified image. Total twelve date NDVI based signature derived from sentinel – 2 data during rabi season 2020 - 2021. NDVI based data set is used to classify the study area's major crops, which are maize, wheat, castor, chilli, cotton, pigeon pea, sorghum, and tobacco. The total research area's growing crops was estimated as 34472.59 ha, 16517.09 ha, 2186.92 ha, 250.59 ha, 8005.56 ha, 1130.93 ha, 1719.38 ha and 1468.52 ha for maize, wheat, castor, chilli, cotton, pigeon pea, sorghum and tobacco, respectively. The overall accuracy and kappa coefficient for crop acreage estimation were calculated to be 90.71% and 0.78%, respectively. The resulted acreage estimation will help to understand the cropping pattern and their interaction with spatial and temporal variability for present and future estimation of crop water requirement and proper resource availability in this selected region.

Fine-Grained Climate Classification for the Qaidam Basin

The Qaidam Basin is a sensitive climate transition zone revealing a wide spectrum of local climates and their variability. In order to obtain an objective and quantitative expression of local climate regions as well as avoid the challenge to pre-define the number of heterogeneous local climates, the ISODATA cluster method is employed to achieve the fine-grained climate divisions of the Qaidam Basin, which can heuristically alter the number of clusters based on the input of monthly temperature and precipitation data. The fine-grained climate classification extends the traditional Köppen climate classification and represents the complex climate transformation processes in terms of fine-grained climate clusters. The following results are observed: (i) The Qaidam Basin is divided into an arid desert basin area and the surrounding alpine mountainous areas. The climate distribution is affected by both the altitude and the dryness ratio, which, employing the Budyko framework, largely characterizes the local energy–water fluxes at the surface and the related vegetation regimes (biomes). The fine-grained climate classification successfully captures their causal relationships and represents them well by the local climates: the climatic spatial differentiation in the mountainous areas is highly consistent with the topography and reveals an elevation-dependent circular distribution from the edges to the center of the basin; the climate heterogeneity within the basin presents a west-to-east meridional distribution due to the combined effect of the mid-latitude westerlies and the Indian monsoon. (ii) The climate gradients are spatially different over the Qaidam Basin. The surrounding mountainous areas have a large climate gradient compared to the inner basin; the southern mountain edge is governed by a more severe climate change than the north-eastern one; and the climate gradient is larger in the eastern than in the western basin. (iii) The lake regions within the basin show an obvious lake effect and reveal a local lake climate. Spatially, a common structure emerges with a dryer-climate zone or watershed embedding a wetter lake-affected area, which appears to migrate eastward becoming stepwise wetter from the very dry center to the wet eastern boundary of the Qaidam basin. This provides a topographically induced insight of the wet climate expansion of initially arid climates and is crucial to improve the Qaidam Basin’s ecological environment. Finally, although this work mainly focuses on the local-scale climates and their variability in the Qaidam Basin, the data-driven cluster methodology for climate refinement is transferable to regional- even global-scale climate studies, which offers broad application prospects.

Detecting wetland encroachment and urban agriculture land classification in Uganda using hyper-temporal remote sensing.

Background: Urbanization is an important indicator of economic growth and social change but is associated with environmental degradation, which threatens the sustainable growth of African cities. One of the most vulnerable ecosystems in urban areas are wetlands. In Uganda, wetlands cover an area of 11% of the country's land area. Half of the wetland areas in Ugandan cities have been converted to industrial and residential areas, and urban agriculture. There is limited information on the extent of wetland conversion or utilization for urban agriculture. The objective of this study was to investigate the extent of wetlands lost in two Ugandan cities, Wakiso and Kampala, in the last 30 years. Secondly, we extracted crop agriculture in the wetlands of Kampala and Wakiso from hyper-temporal satellite image analysis in an attempt to produce a spatial detail of wetland encroachment maps of urban agriculture using a reproducible mapmaking method. Methods: Using a field survey and free remote sensing data from Landsat TM 1986 and Landsat ETM 2016 we classified the rate of wetland loss and encroachment between the years 1986 and 2016. We used MODIS NDVI 16-day composites at a 500-meter spatial resolution to broaden the analysis to distinguish distinctive crops and crop mixtures in the encroached wetlands for urban agriculture using the ISODATA clustering algorithm. Results: Over 30 years, 72,828 ha (73%) of the Wakiso-Kampala wetlands have been lost meanwhile agriculture areas have doubled. Of this 16,488 ha (23%) were converted from wetlands. All cultivated agriculture in Kampala was in the wetlands while in Wakiso, 73% of crop agriculture was in the wetlands. The major crops grown in these urban wetlands were banana (20%), sugarcane (22%), maize (17%), Eucalyptus trees (12%), sweet potatoes (10%), while ornamental nurseries, pine trees, vegetables, and passion fruits were each at 5%.

Delineation of Nitrate Reduction Hotspots in Artificially Drained Areas through Assessment of Small-Scale Spatial Variability of Electrical Conductivity Data.

Identification of nitrate reduction hotspots (NRH) can be instrumental in implementing targeted strategies for reducing nitrate loading from agriculture. In this study, we aimed to delineate possible NRH areas from soil depths of 80 to 180 cm in an artificially drained catchment by utilizing electrical conductivity (EC) values derived by the inversion of apparent electrical conductivity data measured by an electromagnetic induction instrument. The NRH areas were derived from the subzones generated from clustering the EC values via two methods, unsupervised ISODATA clustering and the Optimized Hot Spot Analysis, that highly complement each other. The clustering of EC values generated three classes, wherein the classes with high EC values correspond to NRH areas as indicated by their low redox potential values and nitrate (NO3−) concentrations. Nitrate concentrations in the NRH were equal to 13 to 17% of the concentrations in non-NRH areas and occupied 26% of the total area of the drainage catchments in the study. It is likely that, with the identification of NRH areas, the degree of nitrogen reduction in the vadose zone may be higher than initially estimated at the subcatchment scale.

Fodder crop estimation using Sentinel-2A/B satellite data for West Bengal, India

Mapping, monitoring and estimation of fodder crop area can support farmers, policy makers to make right decisions in various conditions to improve the livestock productivity. The freely available high-resolution Sentinel-2 MSI data has increased the application of precision agriculture for a broader range. This study investigated the capabilities of Sentinel-2A/B MSI satellite data to identify and discriminate the fodder crops from other crops and estimate the area utilized for fodder cultivation in five districts of West Bengal during 2019–20 in rabi season (October to March). Multi-date NDVI based spectral profiles were generated from satellite imagery to ascertain the time of sowing which plays an important role to differentiate fodder crops from other crops. Image classification has been done through the ISODATA clustering technique. A field survey has been carried out in the study area and a total of 610 GT points were collected from different locations of fodder crops using handheld GPS and smartphone-based applications to validate the classification of crops. Accuracy assessment has also been performed between four classes (Fodder crops, Other crops, Forest and Waterbody) which is 81.61% for fodder crops, and 88.97% overall accuracy was observed in classification. The area under fodder cultivation was 0.68% of the total agricultural area, which has been estimated to the tune of 6.78 per thousand hectares. This study can be implemented for fodder crop identification and cultivation management plan to improve the livestock productivity by supply of fodder throughout the year.

Improved ISODATA Clustering Method with Parameter Estimation based on Genetic Algorithm

Improved ISODATA Clustering Method with Parameter Estimation based on Genetic Algorithm

Evaluating land cover types from Landsat TM using SAGA GIS for vegetation mapping based on ISODATA and K-means clustering

Evaluating land cover types from Landsat TM using SAGA GIS for vegetation mapping based on ISODATA and K-means clustering

Mitral Valve Leaflets segmentation approaches based upon Frangi filter and ISODATA clustering

ABSTRACT Early detection of chordae tendineae state through medical imaging has a significant impact in reducing the clinical progression of mitral valve disease. This study proposes two methods for tracking these structures. The first method is based on Frangi filter to detect tubular objects while the second one is mainly based on ISODATA clustering approaches. Twenty-eight patients who had previously undergone CT study for mitral valve evaluation were retrospectively enrolled in the study. Pearson correlation coefficients and a linear regression analysis were used to evaluate the correlation between clinical indices measurements for mitral valve segmentation derived from manual segmentation and those obtained through two semi-automated methods. The results showed that MV segmentation based on Frangi filter gave an average Dice index of 0.79 ± 1.52 and a HD of 2.67± 1.14 mm while the ISODATA algorithm indicated an average Dice index of 0.65 ± 0.24 and a HD of 3.26± 0.95 mm. The Pearson correlation coefficients showed a strong correlation between clinical measurements (r= 0.921 for P-A diameter and r=0.987 for MA area, p < 0.01) for mitral valve segmentation derived from manual segmentation and those obtained through Frangi filter method. Similarly, we reported a good correlation between ISDATA and manual segmentation in the measurement of P-A diameter and MA area with respectively r = 0.872 and r = 0.859. The Frangi filter algorithm performs better than the ISODATA algorithm. This suggests that MV segmentation using Frangi filter algorithm gives a performance, which is quite similar to the expert human segmentation.

An adaptive learning method of anchor shape priors for biological cells detection and segmentation

Background and objective: Owing to the variable shapes, large size difference, uneven grayscale and dense distribution among biological cells in an image, it is still a challenging task for the standard Mask R-CNN to accurately detect and segment cells. Especially, the state-of-the-art anchor-based methods fail to generate the anchors of sufficient scales effectively according to the various sizes and shapes of cells, thereby hardly covering all scales of cells.Methods: We propose an adaptive approach to learn the anchor shape priors from data samples, and the aspect ratios and the number of anchor boxes can be dynamically adjusted by using ISODATA clustering algorithm instead of human prior knowledge in this work. To solve the identification difficulties for small objects owing to the multiple down-samplings in a deep learning-based method, a densification strategy of candidate anchors is presented to enhance the effects of identifying tinny size cells. Finally, a series of comparative experiments are conducted on various datasets to select appropriate a network structure and verify the effectiveness of the proposed methods.Results: The results show that the ResNet-50-FPN combining the ISODATA method and densification strategy can obtain better performance than other methods in multiple metrics (including AP, Precision, Recall, Dice and PQ) for various biological cell datasets, such as U373, GoTW1, SIM+ and T24.Conclusions: Our adaptive algorithm could effectively learn the anchor shape priors from the various sizes and shapes of cells. It is promising and encouraging for a real-world anchor-based detection and segmentation application of biomedical engineering in the future.