top of page

Image Processing Project​

for EECS 351: Intro to Digital Signal Processing

Paul Domanico, Colin Wagner, Jihong Wang, Frank Zhong 

Overview

We are University of Michigan Students in the Digital Signal Processing course, EECS 351. This website is designed to present our work in designing an algorithm to classify satellite imagery. The goal of the algorithm is to input raw satellite imagery and a known road map database into software that can extract features from the imagery and correctly locate its position on the road map.

​

The basis for this project has many applications.  Consider a tool like Google Maps which provides users with interchangeable views of road maps and satellite imagery.  These two views are perfectly aligned to allow for a smooth transition and overlaying of information.  Furthermore, our approaches can be generalized and be considered in applications for the automatic classification of large datasets from remote sensing.

How it Works

We can divide our approach to this task into three main tasks:

  1. Process the road map image for simplification and display of only relevant features

  2. Process the satellite image to pull out or emphasize features most similar to those on the road map

  3. Correlate the two processed outputs to precisely determine where they best fit together

The flow chart below shows this process.

Satellite Image

Road Map

Satellite Image Process

Road Map Process

Correlation Algorithm

Output: Satellite image in most probable location

bottom of page