top of page
Session 14 : Using Python & SAR Images for change detection & deep learning for object recognition handled by Col. Jai Govind, Director NCC - Government of India
Highlights of the Session - 14
• The topics such as environmental remote sensing from security perspective, VEDAS & Future SAR information content from NISAR, Innovative products, SDG 16, World conflicts and resolution, Environmental background, NISAR, QGIS and citizen scientist.
• Basics of Earth Observation data science like Linear algebra, statistics and programming were briefed. He has explained about looking data as vectors, models as processes, learning as iterations in the procedures of AI, Machine and Deep learning.
• Various dimensions of machine learning like supervised, unsupervised, reinforced learning and recommender system for it were explained in detail.
• Insight on the use of python in earth observation and important python libraries like Numpy, Pandas, Matplotlib, Scipy, Scikit learn, Folium/ Geemap, GDAL, keras. Analysis of satellite data by using Numpy was illustrated.
• The differences between visual dimensions and SAR images were narrated and advantages of SAR like penetration through clouds, high change detection capability because of own source, working ability in all-weather conditions, etc were listed.
• Geometry and resolution of SAR was explained in detail and the phenomenon of SAR backscatter depending on radar parameters like frequency/wavelength, polarisation, incidence angle and surface parameters like dielectric constant, surface roughness relative to wavelength, structure & orientation of the surface was also narrated.
• NISAR- the collaboration of NASA and ISRO, in progress was introduced to the gathering and its features, advantages were highlighted.
• SAR techniques like Polarimetric SAR (PolSAR), Interferometric SAR (InSAR), Differential InSAR (DInSAR), small baseline subset (SBAS), Persistent scatter InSAR (PSInSAR), Polarimetric Inferometric SAR (PolInSAR), Tomography.
• The different steps of crisp DM model such as Business understanding, Data understanding, Data preparation, modelling, evaluation, deployment were explained in brief.
• SAR processing techniques involving Focussing, Multilook, CO-registration, Geometric correction, radiometric correction, Speckle filters were narrated in detail.
• SAR Simulation Notebook was illustrated and the detailed explanation on simulation for focusing, geometrical distortions, polarization, Pauli decomposition of scattering, scattering matrix, H-APHA decomposition, Entropy associated with APHA, were given with appropriate examples.
• The process of change detection was defined and the different change detection problems like binary change detection, multiclass change detection, and change detection in long time series of images were narrated with suitable examples.
• The case study of Visual change detection in NGARI GUNSA airport was addressed and the visual versus SAR change detection in this case was explained very well with change detection codes using Google Earth, Folium library, GEOJSON, Sentinel 1 data, etc.
• Per pixel likelihood ratio test for changes between two images based on complex wishart distribution in detail and also about object identification using deep learning and deep learning codes for the same involving keras, flattening the data, transfer learning, etc.
• He also spoke about evaluation in machine learning, object recognition using deep learning including Precision, Recall, F1score, ROC curve, confusion matrix, data simulation and augmentation, etc.
Session 15 : Modern toolkits and advance techniques in ocean colour remote sensing for marine ecosystem studies handled by Dr Arvind Sahay, Scientist E , SAC -ISRO, Government of India
Highlights of the Session - 15
PART-I : WHAT IS OCEAN COLOUR REMOTE SENSING?
• Remote Sensing of ocean colour –
☆ Satellites/ Aerial flights are tools covering large area in very short time and also facilitates repeated observations.
☆ Ocean colour is determined by interaction of incident light with the constituents present in water such as phytoplankton, inorganic particulate matter, CDOM/detritus.
☆ Ocean waters can be put under: case 1 waters (clear water), case 2 waters (slightly turbid), (case 3) optically complex water. Spectral images of case1 and 2 waters were shown and explained.
• Basic principle of ocean colour Remote Sensing.
☆ Figures and formulae correlating reflectance, absorption, scattering, fluorescence, backscattering, water leaving radiance, downwelling irradiated was explained.
• Ocean colour parameter retrieval from space:
☆ It is a 2 step approach comprising of :
1. Atmospheric correction for visible spectral bands – molecular scattering, aerosol scattering.
2. Bio-optical algorithms for water constituent’s estimation using atmospherically correlated radiance.
Atmospheric correction procedures and different global algorithms for phytoplankton’s chlorophyll-a, etc were explained.
• IOP and it’s importance:
IOP-Inherent Optical Properties. Absorption and scattering property ( True characteristic property of ocean constituents like phytoplankton Absorption, CDOM Absorption, backscattering by larger particles in coastal oceans since in coastal oceans, waters are complex.
• IOP studies in Indian waters:
☆ Particle backscattering from water (Gujarat)
☆ Total Absorption of ocean waters in Arabian sea.
☆ CDOM absorption in Indian coastal waters and from hyperspectral data.
All these were elaborated with many case studies and examples.
• Algal blooms from ocean colour measurements:
☆ Definition of bloom, HAB and a video on it, Non-bloom and bloom spectra, different ways to estimate bloom concentrations, etc. with many live demonstrations.
• Phytoplankton size classes (from space):
Micro, Pico and nano planktons.
PART – II: ADVANCE TECHNIQUES
• Analytical inversion techniques- QAA illustration.
• Spectral matching techniques – elaborated with papers published.
• Optimization techniques – explained with published papers.
PART – III: MODERN TOOLKITS
• SeaDAS and SNAP for ocean colour data analysis.
☆ SeaDAS and it’s usage.
☆ Ocean colour monitor data formats(SMI, HDF, etc)
☆ Opening different level (L2 & l2) files and reading the data.
☆ Data Visualization.
☆ Subsetting/ cropping data with shape file.
☆ Analysing the ROI (Mean, Median) – statistics.
☆ Mosaicking two files.
☆ Imposing vector files.
☆ Band math (Averaging two datasets).
All the usages were explained with live demonstration of SeaDAS software.
bottom of page