ajunlonglive / orthomosaic

Rectify and stitch images together using multiview geometry.

Orthomosaic Generator

This project creates an orthomap given a dataset of n images and n corresponding camera poses. It is useful as a postprocessing step for data generated by SLAM techniques such as ORB-SLAM to quickly visualize a scene, especially in cases when most of the imagery is nadir or close to nadir.

A video of the mosaic process is below:

Installation

This project was developed and tested using Python 2.7.10 on Ubuntu 15.10. It depends on NumPy 1.8.2 and OpenCV 2.4.11. Install and execute this project using the following commands:

git clone https://github.com/alexhagiopol/orthomosaic.git
cd orthomosaic
mkdir results  # location where program places results
mkdir datasets  # location where you place input data
sudo apt-get update
sudo apt-get install python-numpy python-opencv
python ImageMosaic.py

Example Dataset

I provide an example "datasets" directory with images and camera poses. You can use this datasets directory instead of creating your own as listed in the instructions above:

wget -O datasets.tar.gz "https://www.dropbox.com/s/3te1zux076f6bwn/datasets.tar.gz?dl=1"
tar -xvzf datasets.tar.gz