Use of multispectral remote sensing data to map magnetite bodies in the Bushveld Complex, South Africa: a case study of Roossenekal, Limpopo.

2021 ◽  
Author(s):  
Mthokozisi Twala ◽  
James Roberts ◽  
Cilence Munghemezulu
Keyword(s):  
Remote Sensing ◽  
Maximum Likelihood ◽  
Bushveld Complex ◽  
The Other ◽  
Support Vector ◽  
Classification Algorithms ◽  
Error Matrix ◽  
Geological Processes ◽  
Eastern Limb ◽  
Multispectral Sensors

<p>The use of remote sensing in mineral detection and lithological mapping has become a generally accepted augmentative tool in exploration. With the advent of multispectral sensors (e.g. ASTER, Landsat, Sentinel and PlanetScope) having suitable wavelength coverage and bands in the Shortwave Infrared (SWIR) and Thermal Infrared (TIR) regions, multispectral sensors have become increasingly efficient at routine lithological discrimination and mineral potential mapping. It is with this paradigm in mind that this project sought to evaluate and discuss the detection and mapping of vanadium bearing magnetite, found in discordant bodies and magnetite layers, on the Eastern Limb of the Bushveld Complex. The Bushveld Complex hosts the world’s largest resource of high-grade primary vanadium in magnetitite layers, so the wide distribution of magnetite, its economic importance, and its potential as an indicator of many important geological processes warranted the delineation of magnetite.</p><p> </p><p>The detection and mapping of the vanadium bearing magnetite was evaluated using specialized traditional, and advanced machine learning algorithms. Prior to this study, few studies had looked at the detection and exploration of magnetite using remote sensing, despite remote sensing tools having been regularly applied to diverse aspects of geosciences. Maximum Likelihood, Minimum Distance to Means, Artificial Neural Networks, Support Vector Machine classification algorithms were assessed for their respective ability to detect and map magnetite using the PlanetScope data in ENVI, QGIS, and Python. For each classification algorithm, a thematic landcover map was attained and the accuracy assessed using an error matrix, depicting the user's and producer's accuracies, as well as kappa statistics.</p><p> </p><p>The Maximum Likelihood Classifier significantly outperformed the other techniques, achieving an overall classification accuracy of 84.58% and an overall kappa value of 0.79. Magnetite was accurately discriminated from the other thematic landcover classes with a user’s accuracy of 76.41% and a producer’s accuracy of 88.66%. The erroneous classification of some mining activity pixels as magnetite in the Maximum Likelihood was inherent to all classification algorithms. The overall results of this study illustrated that remote sensing techniques are effective instruments for geological mapping and mineral investigation, especially in iron oxide mineralization in the Eastern Limb of Bushveld Complex. </p><p> </p>

Detection of magnetite in the Roossenekal area of the Eastern Bushveld Complex, South Africa, using multispectral remote sensing data

2020 ◽  
Vol 123 (4) ◽  
pp. 573-586
Author(s):  
M. Twala ◽  
R. J. Roberts ◽  
C. Munghemezulu
Keyword(s):  
Remote Sensing ◽  
Support Vector Machine ◽  
Maximum Likelihood ◽  
Bushveld Complex ◽  
Support Vector ◽  
Classification Algorithms ◽  
Error Matrix ◽  
Geological Processes ◽  
Multispectral Remote Sensing ◽  
Eastern Limb

Abstract Multispectral sensors, along with common and advanced algorithms, have become efficient tools for routine lithological discrimination and mineral potential mapping. It is with this paradigm in mind that this paper sought to evaluate and discuss the detection and mapping of magnetite on the Eastern Limb of the Bushveld Complex, using high spectral resolution multispectral remote sensing imagery and GIS techniques. Despite the wide distribution of magnetite, its economic importance, and its potential as an indicator of many important geological processes, not many studies had looked at the detection and exploration of magnetite using remote sensing in this region. The Maximum Likelihood and Support Vector Machine classification algorithms were assessed for their respective ability to detect and map magnetite using the PlanetScope Analytic data. A K-fold cross-validation analysis was used to measure the performance of the training as well as the test data. For each classification algorithm, a thematic landcover map was created and an error matrix, depicting the user’s and producer’s accuracies as well as kappa statistics, was derived. A pairwise comparison test of the image classification algorithms was conducted to determine whether the two classification algorithms were significantly different from each other. The Maximum Likelihood Classifier significantly outperformed the Support Vector Machine algorithm, achieving an overall classification accuracy of 84.58% and an overall kappa value of 0.79. Magnetite was accurately discriminated from the other thematic landcover classes with a user’s accuracy of 76.41% and a producer’s accuracy of 88.66%. The overall results of this study illustrated that remote sensing techniques are effective instruments for geological mapping and mineral investigation, especially iron oxide mineralization in the Eastern Limb of the Bushveld Complex.

scholarly journals A Comparative Study of Support Vector Machine and Maximum Likelihood Classification to Extract Land Cover of Lahore District, Punjab, Pakistan

2021 ◽  
Vol 64 (3) ◽  
pp. 265-274
Author(s):  
Fatima Mushtaq ◽  
Khalid Mahmood ◽  
Mohammad Chaudhry Hamid ◽  
Rahat Tufail
Keyword(s):  
Support Vector Machine ◽  
Maximum Likelihood ◽  
Land Cover ◽  
Support Vector ◽  
Kappa Statistics ◽  
Error Matrix ◽  
Classification Techniques ◽  
Machine Learning Classification ◽  
Hyper Plane ◽  
Better Than

The advent of technological era, the scientists and researchers develop machine learning classification techniques to classify land cover accurately. Researches prove that these classification techniques perform better than previous traditional techniques. In this research main objective is to identify suitable land cover classification method to extract land cover information of Lahore district. Two supervised classification techniques i.e., Maximum Likelihood Classifier (MLC) (based on neighbourhood function) and Support Vector Machine (SVM) (based on optimal hyper-plane function) are compared by using Sentinel-2 data. For this optimization, four land cover classes have been selected. Field based training samples have been collected and prepared through a survey of the study area at four spatial levels. Accuracy for each of the classifier has been assessed using error matrix and kappa statistics. Results show that SVM performs better than MLC. Overall accuracies of SVM and MLC are 95.20% and 88.80% whereas their kappa co-efficient are 0.93 and 0.84 respectively.  

scholarly journals Landslide Scars Detection using Remote Sensing and Pattern Recognition Techniques: Comparison Among Artificial Neural Networks, Gaussian Maximum Likelihood, Random Forest, and Support Vector Machine Classifiers

2020 ◽  
Vol 72 (4) ◽  
pp. 665-680
Author(s):  
Tatiana Dias Tardelli Uehara ◽  
Sabrina Paes Leme Passos Corrêa ◽  
Renata Pacheco Quevedo ◽  
Thales Sehn Körting ◽  
Luciano Vieira Dutra ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Pattern Recognition ◽  
Support Vector Machine ◽  
Maximum Likelihood ◽  
Random Forest ◽  
Risk Mitigation ◽  
Vegetation Indices ◽  
Support Vector ◽  
Visual Interpretation ◽  
Artificial Neural

Landslide inventory is an essential tool to support disaster risk mitigation. The inventory is usually obtained via conventional methods, as visual interpretation of remote sensing images, or semi-automatic methods, through pattern recognition. In this study, four classification algorithms are compared to detect landslides scars: Artificial Neural Network (ANN), Maximum Likelihood (ML), Random Forest (RF) and Support Vector Machine (SVM). From Sentinel-2A imagery and SRTM’s Digital Elevation Model (DEM), vegetation indices and slope features were extracted and selected for two areas at the Rolante River Catchment, in Brazil. The classification products showed that the ML and the RF presented superior results with OA values above 92% for both study areas.  These best accuracy’s results were identified in classifications using all attributes as input, so without previous feature selection.

scholarly journals MAPPING OF BURNED AREA USING PRESENCE AND BACKGROUND LEARNING FRAMEWORK ON THE GOOGLE EARTH ENGINE PLATFORM

2019 ◽  
Vol XLII-3/W8 ◽  
pp. 37-41
Author(s):  
D. Attaf ◽  
K. Djerriri ◽  
D. Mansour ◽  
D. Hamdadou
Keyword(s):  
Remote Sensing ◽  
Classification Problem ◽  
Google Earth ◽  
The Other ◽  
Burned Area ◽  
Support Vector ◽  
Classification Techniques ◽  
Burned Areas ◽  
Google Earth Engine ◽  
One Class Classification

<p><strong>Abstract.</strong> Mapping of burned areas caused by forest fires was always a main concern to researchers in the field of remote sensing. Thus, various spectral indices and classification techniques have been proposed in the literature. In such a problem, only one specific class is of real interest and could be referred to as a one-class classification problem. One-class classification methods are highly desirable for quick mapping of classes of interest. A common used solution to deal with One-Class classification problem is based on oneclass support vector machine (OC-SVM). This method has proved useful in classification of remote sensing images. However, overfitting problem and difficulty in tuning parameters have become the major obstacles for this method. The new Presence and Background Learning (PBL) framework does not require complicated model selection and can generate very high accuracy results. On the other hand the Google Earth Engine (GEE) portal provides access to satellite and other ancillary data, cloud computing, and algorithms for processing large amounts of data with relative ease. Therefore, this study mainly aims to investigate the possibility of using the PBL framework within the GEE platform to extract burned areas from freely available Landsat archive in the year 2015. The quality of the results obtained using PBL framework was assessed using ground truth digitized by qualified technicians and compared to other classification techniques: Thresholding burned area spectral Index (BAI) and OC-SVM classifiers. Experimental results demonstrate that PBL framework for mapping the burned areas shows the higher classification accuracy than the other classifiers, and it highlights the suitability for the cases with few positive labelled samples available, which facilitates the tedious work of manual digitizing.</p>

scholarly journals Development of Novel Classification Algorithms for Detection of Floating Plastic Debris in Coastal Waterbodies Using Multispectral Sentinel-2 Remote Sensing Imagery

2021 ◽  
Vol 13 (8) ◽  
pp. 1598
Author(s):  
Bidroha Basu ◽  
Srikanta Sannigrahi ◽  
Arunima Sarkar Basu ◽  
Francesco Pilla
Keyword(s):  
Remote Sensing ◽  
Near Infrared ◽  
Marine Ecosystem ◽  
Training Data ◽  
Support Vector ◽  
Classification Algorithms ◽  
Real World Data ◽  
Plastic Debris ◽  
Fuzzy C Means ◽  
Sentinel 2

Plastic pollution poses a significant environmental threat to the existence and health of biodiversity and the marine ecosystem. The intrusion of plastic to the food chain is a massive concern for human health. Urbanisation, population growth, and tourism have been identified as major contributors to the growing rate of plastic debris, particularly in waterbodies such as rivers, lakes, seas, and oceans. Over the past decade, many studies have focused on identifying the waterbodies near the coastal regions where a high level of accumulated plastics have been found. This research focused on using high-resolution Sentinel-2 satellite remote sensing images to detect floating plastic debris in coastal waterbodies. Accurate detection of plastic debris can help in deploying appropriate measures to reduce plastics in oceans. Two unsupervised (K-means and fuzzy c-means (FCM)) and two supervised (support vector regression (SVR) and semi-supervised fuzzy c-means (SFCM)) classification algorithms were developed to identify floating plastics. The unsupervised classification algorithms consider the remote sensing data as the sole input to develop the models, while the supervised classifications require in situ information on the presence/absence of floating plastics in selected Sentinel-2 grids for modelling. Data from Cyprus and Greece were considered to calibrate the supervised models and to estimate model efficiency. Out of available multiple bands of Sentinel-2 data, a combination of 6 bands of reflectance data (blue, green, red, red edge 2, near infrared, and short wave infrared 1) and two indices (NDVI and FDI) were selected to develop the models, as they were found to be most efficient for detecting floating plastics. The SVR-based supervised classification has an accuracy in the range of 96.9–98.4%, while that for SFCM and FCM clustering are between 35.7 and 64.3% and 69.8 and 82.2%, respectively, and for K-means, the range varies from 69.8 to 81.4%. It needs to be noted that the total number of grids with floating plastics in real-world data considered in this study is 59, which needs to be increased considerably to improve model performance. Training data from other parts of the world needs to be collected to investigate the performance of the classification algorithms at a global scale.

scholarly journals Evaluación de cuatro algoritmos de clasificación de imágenes satelitales Landsat-8 y Sentinel-2 para la identificación de cobertura boscosa en paisajes altamente fragmentados en Costa Rica

2020 ◽  
pp. 37
Author(s):  
I.D. Ávila-Pérez ◽  
E. Ortiz-Malavassi ◽  
C. Soto-Montoya ◽  
Y. Vargas-Solano ◽  
H. Aguilar-Arias ◽  
...  
Keyword(s):  
Support Vector Machine ◽  
Costa Rica ◽  
Maximum Likelihood ◽  
Forest Cover ◽  
Support Vector ◽  
Classification Algorithms ◽  
Landsat 8 ◽  
Classification Methods ◽  
Global Accuracy ◽  
Sentinel 2

<p>Mapping of land use and forest cover and assessing their changes is essential in the design of strategies to manage and preserve the natural resources of a country, and remote sensing have been extensively used with this purpose. By comparing four classification algorithms and two types of satellite images, the objective of the research was to identify the type of algorithm and satellite image that allows higher global accuracy in the identification of forest cover in highly fragmented landscapes. The study included a treatment arrangement with three factors and six randomly selected blocks within the Huetar Norte Zone in Costa Rica. More accurate results were obtained for classifications based on Sentinel-2 images compared to Landsat-8 images. The best classification algorithms were Maximum Likelihood, Support Vector Machine or Neural Networks, and they yield better results than Minimum Distance Classification. There was no interaction among image type and classification methods, therefore, Sentinel-2 images can be used with any of the three best algorithms, but the best result was the combination of Sentinel-2 and Support Vector Machine. An additional factor included in the study was the image acquisition date. There were significant differences among months during which the image was acquired and an interaction between the classification algorithm and this factor was detected. The best results correspond to images obtained in April, and the lower to September, month that corresponds with the period of higher rainfall in the region studied. The higher global accuracy identifying forest cover is obtained with Sentinel-2 images from the dry season in combination with Maximum Likelihood, Support Vector Machine, and Neural Network image classification methods.</p>

scholarly journals Mapping vegetation communities of the Karkonosze National Park using APEX hyperspectral data and Support Vector Machines

2014 ◽  
Vol 18 (2) ◽  
pp. 23-29 ◽  
Author(s):  
Adriana Marcinkowska ◽  
Bogdan Zagajewski ◽  
Adrian Ochtyra ◽  
Anna Jarocińska ◽  
Edwin Raczko ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Support Vector Machines ◽  
National Park ◽  
Kappa Coefficient ◽  
Hyperspectral Data ◽  
Support Vector ◽  
Svm Classifier ◽  
Error Matrix ◽  
Vegetation Communities ◽  
Vector Machines

Abstract This research aims to discover the potential of hyperspectral remote sensing data for mapping mountain vegetation ecosystems. First, the importance of mountain ecosystems to the global system should be stressed due to mountainous ecosystems forming a very sensitive indicator of global climate change. Furthermore, a variety of biotic and abiotic factors influence the spatial distribution of vegetation in the mountains, producing a diverse mosaic leading to high biodiversity. The research area covers the Szrenica Mount region on the border between Poland and the Czech Republic - the most important part of the Western Karkonosze and one of the main areas in the Karkonosze National Park (M&B Reserve of the UNESCO). The APEX hyperspectral data that was classified in this study was acquired on 10th September 2012 by the German Aerospace Center (DLR) in the framework of the EUFAR HyMountEcos project. This airborne scanner is a 288-channel imaging spectrometer operating in the wavelength range 0.4-2.5 μm. For reference patterns of forest and non-forest vegetation, maps (provided by the Polish Karkonosze National Park) were chosen. Terrain recognition was based on field walks with a Trimble GeoXT GPS receiver. It allowed test and validation dominant polygons of 15 classes of vegetation communities to be selected, which were used in the Support Vector Machines (SVM) classification. The SVM classifier is a type of machine used for pattern recognition. The result is a post classification map with statistics (total, user, producer accuracies, kappa coefficient and error matrix). Assessment of the statistics shows that almost all the classes were properly recognised, excluding the fern community. The overall classification accuracy is 79.13% and the kappa coefficient is 0.77. This shows that hyperspectral images and remote sensing methods can be support tools for the identification of the dominant plant communities of mountain areas.

Glioblastoma gene expression subtypes and correlation with clinical, molecular and immunohistochemical characteristics in a homogenously treated cohort: GLIOCAT project.

2019 ◽  
Vol 37 (15_suppl) ◽  
pp. 2029-2029 ◽  
Author(s):  
Estela Pineda ◽  
Anna Esteve-Codina ◽  
Maria Martinez-Garcia ◽  
Francesc Alameda ◽  
Cristina Carrato ◽  
...  
Keyword(s):  
Gene Expression ◽  
Web Application ◽  
Nearest Neighbor ◽  
Expression Profiles ◽  
Gene Set Enrichment Analysis ◽  
The Other ◽  
Support Vector ◽  
Classification Algorithms ◽  
K Nearest Neighbor ◽  
Tissue Samples

2029 Background: Glioblastoma (GBM) gene expression subtypes have been described in last years, data in homogeneously treated patients is lacking. Methods: Clinical, molecular and immunohistochemistry (IHC) analysis from patients with newly diagnosed GBM homogeneously treated with standard radiochemotherapy were studied. Samples were classified based on the expression profiles into three different subtypes (classical, mesenchymal, proneural) using Support Vector Machine (SVM), the K-nearest neighbor (K-NN) and the single sample Gene Set Enrichment Analysis (ssGSEA) classification algorithms provided by GlioVis web application. Results: GLIOCAT Project recruited 432 patients from 6 catalan institutions, all of whom received standard first-line treatment (2004 -2015). Best paraffin tissue samples were selected for RNAseq and reliable data were obtained from 124. 82 cases (66%) were classified into the same subtype by all three classification algorithms. SVM and ssGEA algorithms obtain more similar results (87%). No differences in clinical variables were found between the 3 GBM subtypes. Proneural subtype was enriched with IDH1 mutated and G-CIMP positive tumors. Mesenchymal subtype (SVM) was enriched in unmethylated MGMT tumors (p = 0.008), and classical (SVM) in methylated MGMT tumors (p = 0.008). Long survivors ( > 30 months) were rarely classified as mesenchymal (0-7.5%) and were more frequently classified as Proneural (23.1-26.). Clinical (age, resection, KPS) and molecular ( IDH1, MGMT) known prognostic factors were confirmed in this serie. Overall, no differences in prognosis were observed between 3 subtypes, but a trend to worse survival in mesenchymal was observed in K-NN (9.6 vs 15 ). Mesenchymal subtype presented less expression of Olig2 (p < 0.001) and SOX2 (p = 0.003) by IHC, but more YLK-40 expression (p = 0.023, SVM). On the other hand, classical subtype expressed more Nestin (p = 0.004) compared to the other subtypes (K-NN). Conclusions: In our study we have not found correlation between glioblastoma expression subtype and outcome. This large serie provides reproducible data regarding clinical-molecular-immunohistochemistry features of glioblastoma genetic subtypes.

Sign in / Sign up

Export Citation Format

Share Document