Research Projects - BITS Pilani

Gopal Singh Phartiyal

Assistant Professor

Earth Observation Imagery and Data Processing, Geospatial Data Science, PolSAR/InSAR Image Processing

Department of Computer Science & Information Systems,
Birla Institute of Technology & Science, Pilani- 333031, Rajasthan. India.

gopal.phartiyal@pilani.bits-pilani.ac.in

+91 -1596-255730

+91-7417931696

1. DEEPVOLC: Forecasting volcanic activity using deep learning (DEEPVOLC) | University of Leeds, Feb 5, 2024 – Till date

Development of the latest deep learning algorithms to the satellite data to combine knowledge from all volcanoes that have been active in the satellite-monitoring era. This enable to use the knowledge of how volcanoes behave globally to identify deformation at volcanoes locally, and forecast how it will evolve.
To develop deep learning approaches to detect, classify, and forecast surface deformation at volcanoes, based on InSAR data from radar satellites. To use data on how volcanoes, behave globally to automatically identify and forecast, deformation at volcanoes locally.
Adapting deep neural networks to take additional inputs, such as elevation and weather models, to reduce noise in deformation outputs and improve forecasts.
Developing machine learning approaches to estimate parameters for modeled physical processes causing volcano deformation;

2. Development of Auto-Change Detection System for Pipeline RoU Monitoring using Deep Learning, WebGIS, and VHR Satellite Images

| IIT Roorkee, Apr. 22, 2022 – Till date

Working with very high-resolution multispectral satellite imagery (< 1meter)
Developing object-based image analysis and deep learning algorithms for change mapping.
Deploying the analytics and processing on WebGIS-based application.
Development of object-based change detection methodology for RoU surveillance.
Development of deep learning models for object-based change detection with very high-resolution imagery.
Integrating geospatial data-processing tools
Developing GIS-based Web-UI for data visualization and deploying processed imagery on the web.
Administrative: Project management, reporting, and user manuals, team management, very high-resolution (< 1meter) satellite imagery data purchase

Development of Novel Deep learning approaches for the extraction of information from satellite images | IIT Roorkee, July 2015 – Jun 2021

Analysis and development of novel deep learning and other advanced deep learning and machine learning approaches for various satellite image based applications such as earth surface monitoring, disaster monitoring, missing information reconstruction, agriculture crop monitoring. Novel convolutional and recurrent neural networks (CNNs and RNNs) are proposed and developed under the project.

Optimal Inference in Complex and Turbulent Data | IIT Roorkee, INRIA, Bordeaux, France, March 19, 2015- August 31, 2017

Optimal Information estimation from polarimetric synthetic aperture radar or PolSAR image data in the presence of speckle noise due to layover, foreshortening, or other phenomena. The project involved development of adaptive and optimal class boundary estimation algorithm using statistics present in the image data and evolutionary optimization algorithms.

Comparative assessment of various Ambiguity Resolution techniques in GPS data | IIT Roorkee, July 2012 - June 2013

The project includes reading receiver independent GPS data exchange or RINEX format, correcting time implication, Standard points positing, and computing baseline correction, and ambiguity resolution. Implementation of the LAMBDA ambiguity resolution method in Matlab

Integrated first degree