Orbitia - Change Detection & Segmentation in Remote Sensing

This project focuses on using ML Techniques and deep learning models to better analyse satellite imagery in remote sensing.

Description

The purpose of this project is to use satellite images of a place of different timestamps and finding areas that changed over time. The changes can also be classified into different objects like roads, buildings, etc. As satellite imagery users multiple bands to store data(other than RGB bands), the application allows you to analyse multiple characteristics of the region like vegetation, moisture, geology, etc using combinations of other bands like infrared, SWIR, etc.

Techniques & Algorithms

• Principal Component Analysis
• KMeans
• Machine learning
• Deep learning
• UNET Network
• Data Pre-Processing
• Data Augmentation

Techologies

• Python
• Pandas, Sklearn, Jupyter
• Flask
• HTML, CSS, JQuery
• Tensorflow

Datasets used

Building Detection and Road Detection datasets provided by Spacenet.ai were used. Spacenet.ai provides these public datasets as a part of challenges that frequently conduct. Datasets can be found at - https://spacenet.ai/datasets/

Demo video link - https://drive.google.com/file/d/1Ft_ow4ph5StnUc2EF8rlBNbe-eINDeyw/view?usp=sharing

Segmentation

Detecting changes