6th International Conference on ICT for Intelligent Systems (ICTIS – 2022)  Application of Machine Learning in Mineral Mapping using Remote Sensing 

[Audio] Welcome to the Presentation on Application of Machine Learning in Mineral Mapping using Remote Sensing.

[Audio] We will cover Study Background Problem Statement Framework Conclusion References.

[Audio] There are two main facets to this Research : Machine Learning & Remote Sensing The Machine Learning is an effective approach towards acquiring patterns. Remote Sensing is the field of study the imageries captured by the satellite may have a resolution of varied types including; spatial, spectral and temporal imagery.

Introduction. Remote Sensing is one popular field that can be employed with the ML concepts With the raw spatial data captured by the sensors like LANDSAT,MODIS meaningful insights can be drawn adhering to the specifics of the arena. The images are captured in the form of electromagnetic waves The paper intends to showcase the relevance of machine learning concepts pertaining to a specific area of application in Geosciences with the identification of potential mineral mapping areas as the key objective. The Sensor Captures Surface Image in the form of spectral Signatures The Surface is analysed for the common data points and categorised based on the application area Larger Area can be covered The spectral imagery can be collected using multispectral or hyperspectral sensors the difference being in the number of bands present in the images captured.

Scope. Application areas of remote sensing akin to the field of lithology with mineral mapping considering little to barren areas subjected to mineral exploration accustomed to pattern recognition and fracture detection employing Machine learning and thereby earth space data science essentials..

Introduction. Several potential regions with mineral deposits have been identified on earth’s surface area and research is being carried out to explore the regions with capacitive mineral bank digging below the surface regions. Even though site mapping and physical scrutiny of geological assets are more effective, the aerial and satellite capture employing remote sensing arena facilitates insight into wider geographical locations as well as capture required data in the format required, especially in the era where technology has become the primary source of aiding solution to the environmental crisis. Remote sensing can be used to collect valuable information in the form of satellite imagery that can be supplemented with Machine Learning processing to draw meaningful insights into discovering potential areas of mineral mapping . The proposed Solution accustoms to develop and design of feasible solution for Multispectral analysis and Geological Mapping employing Earth Observation remote sensing data and Machine Learning Algorithms.

Research Considerations. EMR: Electromagnetic Radiation Atmospheric Scattering Atmospheric Corrections Spectral Signatures Geospatial Data Analysis and Machine Learning Accuracy Assessment of Land Cover.

Vision of the Work. Differentiate between Landcover Find Age Relationships between Rocks Differentiate within the Landcover Analyze Minerals Identify Capacitive Mineral Bank.

Background. The Sensor Captures Surface Image in the form of spectral Signatures The Surface is analysed for the common data points and categorised based on the application area Larger Area can be covered The spectral imagery can be collected using multispectral or hyperspectral sensors the difference being in the number of bands present in the images captured A satellite sensors collects light energy within specific regions of the electromagnetic spectrum. Each region in the spectrum is referred to as a band..

Key Findings. The type of rock can demonstrate the presence of a specific kind of mineral in the region. The environmental conditions affect the rock alterations and thereby spatial associations can also be derived effectively. With the discussed hypothesis we can infer that geological structure and extant rock can imply data points on the mineral deposits to mapped accordingly. The satellite imagery is associated with spectral signatures captured in the form of Electromagnetic waves. For the Minerals and rocks, the spectral signatures can be analyzed to detect anomalies and patterns on the persistence lattice structure. With terabytes of spectral signatures gathered by the sensors , automated information extraction processing and recognition can be supplemented by employing data mining approaches , thereby delving into implementation of suitable Machine Learning Algorithms. Improved data collection with limited anomalies can generate desired results. Post data collection applying remote sensing concepts, gathering data from sensors like LANDSAT and MODIS , the preprocessing of data is the next requisite. Principle Component Analysis (PCA) can be performed on the geospatial images for exploration of mineral deposits. The visual appearances of the rocks are essential data captured by the sensors that can draw insights on the associated mineral deposits. The texture and colors of the rocks when captures by the sensors are absorbed at different wavelength. To comprehend the relationship between lithology and mineral deposits, let us have an association formulation. The rocks of the classification like Amphibole, Biotite, Olivine, Plagiocase , Pyroxine , Quartz, etc differ in their color, fracture, form and structure. They vary ranging from being crystalline, flaky, and granular to being irregular grains. Exemplifying the association further Biotite mica is a part of the felsic mineral group and dark color mica are called the biotite. Hardness is another scale of measurement for a rock that have evolved over time. The Cleavage of the rock can result as a 2 perfect set or a 1 perfect set. The spectra or the spectral resolution depends on the idiosyncrasy of the different elements. Landsat-7 and Landsat 8 data can be used to identify minerals deposits employing hydrothermal alterations in rocks. Potential mineral mapping regions can be obtained by employing spectral unmixing of hyperspectral data(or multispectral data) based on the spectral signature of mineral. The images from the sensors are captured in the form of Electromagnetic Radations and Electromagnetic Waves with varied intensity pertaining to the specifics of the captured image.

Spectral Unmixing. Layers Map Satellite. Classification Algorithms and Methods.

Band Composition. Tracking the temporal resolution we can identify the alteration in the rock sediments over the years. Geomorphology, folds, fractures and faults determine maximum information from available data. Visible Near Infrared and Short Wave Infrared are powerful tools of geology to capture data points of minerals. The spectral band composition pertaining to intended data set is given.

Band Specifications. Band Band Specifics Landsat 8-OLI Wavelength LT-8 Mapping Band Specifics Landsat 7- ETM + Wavelength LT -7 Mapping Band 1 Coastal Aerosol 0.43-0.45 Coastal and aerosol studies Blue 0.45-0.52 Differentiate soil from vegetation and deciduous from coniferous vegetation Band 2 Blue 0.45-0.51 Differentiation between soil from vegetation and deciduous from coniferous vegetation Green 0.52-0.60 Emphasizes peak vegetation Band 3 Green 0.53-0.59 Emphasizes peak vegetation Red 0.63-0.69 Discriminates vegetation slopes Band 4 Red 0.64-0.67 Discriminates vegetation slopes NIR 0.77-0.90 Focuses on biomass content and shorelines Band 5 IR 0.85-0.88 Focuses on biomass content and shorelines SWIR 1.55-1.75 Discriminates moisture content of soil and vegetation Band 6 SWIR 1 1.57-1.65 Discriminates moisture content of soil and vegetation; penetrates thin clouds TIR 10.40-12.50 Thermal mapping and predicted soil moisture Band 7 SWIR 2 2.11-2.29 Improved moisture content of soil and vegetation SWIR 2.09-2.35 Hydrothermally altered rocks associated with mineral deposits Band 8 Panchromatic 0.50-0.68 Sharper image definition Panchromatic 0.52-0.90 Sharper image definition Band 9 Cirrus 1.36-1.38 Improved detection of cirrus cloud contamination Band 10 TIRS 1 10.60-11.19 Thermal mapping and predicted soil moisture Band 11 TIRS 2 11.50-12.51 Improved Thermal mapping and estimated soil moisture.

Geospatial Analysis using Machine Learning. Google Earth Geospatial Data Analysis can be performed on cloud employing Google Earth Engine, Landsat 8- Operational Land Imager (OLI) and Landsat‐7 Enhanced Thematic Mapper plus (ETM+) can be used. The satellite has a repository of earth’s images with a 16- Day repetition cycle. Considering portions of India on Geographical maps, we select the target area by marking the polygon on GEE or importing the location points. To limit the data acquisition with less than 5000 records, filter bounds are applied to the Landsat 8-OLI dataset. As the target area pertains top locations in India, the filter is ap-plied to the country considered. The properties and features of the Landsat-8 data can be acquired by printing the same post application of filterBounds . Different combinations of spectral Bands to comprehend the spectral reflectance..

Geospatial Analysis using Machine Learning. Google Earth Geospatial Data Analysis can be performed on cloud employing Google Earth Engine, Landsat 8- Operational Land Imager (OLI) and Landsat‐7 Enhanced Thematic Mapper plus (ETM+) can be used. The satellite has a repository of earth’s images with a 16- Day repetition cycle. Considering portions of India on Geographical maps, we select the target area by marking the polygon on GEE or importing the location points. To limit the data acquisition with less than 5000 records, filter bounds are applied to the Landsat 8-OLI dataset. As the target area pertains top locations in India, the filter is ap-plied to the country considered. The properties and features of the Landsat-8 data can be acquired by printing the same post application of filterBounds . Different combinations of spectral Bands to comprehend the spectral reflectance..

Machine Learning Facet. The supervised image classification process adopting suitable algorithm is further processed. Defining the land cover classes separated as spectral bands, the training site is selected. The data points are then generalized as statistical parameters. The accuracy of the adopted algorithm is assessed appropriated by suitable measure discussed further in the section. The accuracy Assessment stage is followed by the output stage of image classification..

Geospatial Analysis using Machine Learning. the 7 RTM+ LANDSAT S OLI)in GEE code Editor Filter for Ccxvdinates(Select the Filter data by Date geometry Polygon on GEE) Set the selection Bands in combination as listed In •lable (Bl. B2. 133...) Sampling of the Regions of predictor bands and select our training data) var trainingC1assifier —image. gel ect (predi ctBand3. sampl eRegions collect ion : trainData, properties: ( 'land_C1ass' I Y); Training of the classifiers( Random Forest Supervised Machine Learning Algorithm is adopted, demonstrative) var classifier— ee . Classifier . smileRandomForest (10) . train (features : tr ainingClassifier, classProperty: 'land class, In utPro erties: redictBands) ; Get the classified Image Add the classified Image to the map.

Geospatial Analysis using Machine Learning. hijiii liftiiiii.

Conclusion. The mining sector is poised to play a crucial role in the revival of the economy. It will form the foundation of all our efforts towards the much-needed economic recovery dented by the pandemic. The remote discovery of metallic and non-metallic minerals can pave ways to effective solution for exploration thereby leading to exploitation and reclamation. ML- based approach are widely used in remote sensing-based applications. With the paper proposed, the role of sensors in data collection of multispectral images has been presented and the role of machine learning for feature extraction is discussed. Post collection of images in the form of spectral signatures suitable ML approach can be adopted which is succeeded by accuracy assessment by suitable statistical analysis. The current work intends to showcase the application of Machine Learning facet with satellite imagery collected harnessing the potential of Remote sensing data. The work focuses on presenting an extensive scrutiny of spectral unmixing between the Landcovers. The extension of the work will cover the spectral unmixing within the landcovers and classify the potential mining sites based on lithological data points..

References. Jung, D., & Choi, Y.: Systematic review of machine learning applications in mining: Exploration, exploitation, and reclamation. Minerals, 11(2), 1–20 (2021) Shirmard , H., Farahbakhsh , E., Muller, D., & Chandra, R.: A review of machine learn- ing in processing remote sensing data for mineral exploration. Remote Sensing of Envi- ronment , 0034-4257 (2021) Li, T., Zuo , R., Xiong , Y., & Peng, Y.: Random-Drop Data Augmentation of Deep Convolutional Neural Network for Mineral Prospectivity Mapping. Natural Resources Research, 30(1), 27–38 (2021) Chung, B., Yu, J., Wang, L., Kim, N. H., Lee, B. H., Koh, S., & Lee, S.: Detection of magnesite and associated gangue minerals using hyperspectral remote sensing-a labora -tory approach. Remote Sensing, 12(8) (2020) Bolouki , S. M., Ramazi , H. R., Maghsoudi , A., Pour, A. B., & Sohrabi , G.: A remote sening -based application of bayesian networks for epithermal gold potential mapping in AharArasbaran area, NW Iran. In Remote Sensing 12(1) (2020) Cardoso-Fernandes, J., Teodoro, A. C., Lima, A., & Roda -Robles, E.: Semi-automatization of support vector machines to map lithium (Li) bearing pegmatites. Remote Sensing, 12(14) (2020) Díez -Pastor, J. F., Jorge-Villar, S. E., Arnaiz-González, Á., García-Osorio, C. I., Díaz- Acha , Y., Campeny , M., Bosch, J., & Melgarejo , J. C.: Machine learning algorithms ap-plied to Raman spectra for the identification of variscite originating from the mining complex of Gavà . Journal of Raman Spectroscopy, 51(9), 1563–1574 (2020) Nair, P., Srivastava, D. K., & Bhatnagar, R.: Remote sensing roadmap for mineral map-ping using satellite imagery. 2nd International Conference on Data, Engineering and Applications, IDEA 2020. (2020) Li, S., Chen, J., & Xiang, J.: Applications of deep convolutional neural networks in prospecting prediction based on two-dimensional geological big data. Neural Compu-ting and Applications, 32(7), 2037–2053 (2020), Kaplan, U. E., & Topal , E.: A new ore grade estimation using combine machine learning algorithms. Minerals, 10(10), 1–17.(2020).

References. 11. Rajan Girija , R., & Mayappan , S.: Mapping of mineral resources and lithological units: a review of remote sensing techniques. International Journal of Image and Data Fusion, 10(2), 79–106 (2019) 12. Schnitzler, N., Ross, P. S., & Gloaguen , E.: Using machine learning to estimate a key missing geochemical variable in mining exploration: Application of the Random Forest algorithm to multi-sensor core logging data. Journal of Geochemical Exploration, 205 (2019) 13. Lary , D. J., Alavi , A. H., Gandomi , A. H., & Walker, A. L.: Machine learning in geosci-ences and remote sensing. Geoscience Frontiers, 7(1), 3–10 (2016) 14. Notesco , G., Kopačková , V., Rojík , P., Schwartz, G., Livne , I., & Dor , E. Ben.: Mineral classification of land surface using multispectral LWIR and hyperspectral SWIR remote sensing data. A case study over the sokolov lignite open-pit mines, the Czech Republic. Remote Sensing, 6(8), 7005–7025 (2014) 15. Li, J.: Texture classification of landsat TM imagery using bayes point machine. Pro- ceedings of the Annual Southeast Conference. (2013) 16. Liu, L., Zhou, J., Jiang, D., Zhuang, D., Mansaray, L. R., & Zhang, B.: Targeting miner-al resources with remote sensing and field data in the Xiemisitai area, West Junggar , Xin-jiang, China. Remote Sensing, 5(7), 3156–3171 (2013) 17. Lary , D. J., Remer, L. A., MacNeill, D., Roscoe, B., & Paradise, S.: Machine learning and bias correction of MODIS aerosol optical depth. IEEE Geoscience and Remote Sensing Letters, 6(4), 694–698 (2009) 18. Rajesh, H. M.: Application of remote sensing and GIS in mineral: Resource mapping - An overview. Journal of Mineralogical and Petrological Sciences, 99(3), 83–103 (2004.