Sensing and spatial modeling for earth observation

Image formation

This week we see how a digital image is formed in a perspective camera and what are the basic geometrical relations between image, camera and object frames. We also exercise how to correct image observations for lens distortions. 

Teacher: Jan Skaloud / Devis Tuia

Assistants: Aurélien Brun, Kyriaki Mouzakidou, Jesse La Haye 

Keypoint detection, description and matching

This week we will see how detect salient keypoints in images, describe them in terms of features, and how to use those to find correspondences that can be used for matching.

Teacher: D. Tuia

Assistants: E. Dalsasso, J. Sauder

Image orientation

This week we see how to orient and calibrate an image from known points and practice it with one method using the observations from Ex. 1. We also get acquainted with the basics of depth determination in a stereo-vision system. 

Teacher: Jan Skaloud

Assistants:  Jesse La Haye, Antoine Carreau

Stereo vision

We will first see how to obtain depth from a stereo-pair of oriented images, then how to infer the relative orientation (pose) of two cameras from point-to-point correspondences (key-point matches). This process can generalise to multiple-views (images). 

Teacher: Jan Skaloud

Assistants: Jesse LaHaye, Antoine Carreau

Mapping products: DEM and ortho-photo

In this week we will see how obtain a dense 3D point-cloud from the globally optimised (oriented and calibrated) images and how to transform it to a digital elevation model (DEM). The DEM, together with the oriented photos is then used for creating an ortho-photo (=ortho-image), that similarly to a plan/map has correct distances over the whole scene. 

Teacher: Jan Skaloud

Assistants: Jesse LaHaye, Antoine Carreau

Feature extraction from the DEM

This week we will see how extract spatial features from a digital elevation model (like directional derivatives, slopes, etc) and from images. 

These features will be then re-used in the following weeks to train machine learning algorithms.

Teacher: D. Tuia

Assistants: H. Porta, G. Sümbül

Important: There will be no exercise on Friday. The "features" exercise will be next Friday.

In this short lecture we will discuss what is machine learning and go through the main families of methods with some examples.

On Friday, we will have the exercise on feature extraction (relative to the course you had on the 27.03)

Teacher: D. Tuia

Assistants: H. Porta, E. Dalsasso

Part dedicated to Geostatistics

Teacher: Alexis Berne

Assistants: G. Ghiggi and M. Guidicelli