Satellite Image Change Detection (SICD)

Abstract

SICD is primarily a machine learning based project that include data analysis and data visualization of images (that will be taken from satellite) of different dates. It will have a dashboard, on which there will be an option to upload two images of different dates and a button to select whether to detect roads or buildings from the image. In this project, we will build a Satellite Image Change Detection System which will take two images of different dates as an input and find the changes between two images. Initially in our project we are training our model for detection of roads in two images and show them side by side to see the change related to roads. The system will be deployed on a website where GUI will be displayed for input of two images. Our model will be trained using tensor flow where training, testing and validation of the dataset will be done. The system will take images and will detect roads and display the roads in grey scale image. We will add option for buildings detection in the interface so that in future, anyone can deploy model related to building detection on website after training the dataset. This project is web-based application so we will be able to access the system from anywhere if we have the IP address for that website.