Skip to content

Stereoprocessing

Overview

The module 5 of the course is dedicated to an introduction on the stereo processing from fixed-time-lapse cameras.

Learning outcomes

In this module you will learn:

  • how to perform a feature matching between two images by using Deep Learning algorithms, that are necessary for wide baseline stereo reconstruction;
  • how to perform a stereo reconstruction from a stereo pair of images acquired in a single epoch;
  • how to perform a multitemporal stereo reconstruction from a set of stereo pairs of images acquired in different epochs.

Table of contents

  1. Getting started: exaplin how to install ICEpy4D that will used for the processing.

  2. Data preparation: explains how to download the data and how to organize it for the processing.

  3. Feature matching: explain how to perform a feature matching between two images by using Deep Learning algorithms, that are necessary for wide baseline stereo reconstruction.

  4. Single epoch stereo reconstruction: explain how to perform a stereo reconstruction from a stereo pair of images acquired in a single epoch.

  5. Multitemporal stereo reconstruction: explain how to perform a multitemporal stereo reconstruction from a set of stereo pairs of images acquired in different epochs.

Software

We will use ICEpy4D Python toolkit for the stereo processing. ICEpy4D is an open-source project for multitemporal photogrammetry developed by the Geodesy and Geomatics Laboratory of Politecnico di Milano and it is available on GitHub at https://github.com/franioli/icepy4d.

For the Bundle Adjustment and the dense reconstruction, you also need Agisoft Metashape Professional Edition and the Metashape Python API (see Getting started for more details).

A Linux environment is strongly recommended for the processing, but also Windows will work.