A Generic Framework for Library-based Mapping
of Urban Areas
What is it about?
Cities are growing faster than ever, and so does the amount of earth observation imagery covering urban areas. We are reaching a point where the sheer extent of image databases exceeds our ability to map them into useful information.
Much research has been dedicated to making more advanced mapping applications, fueled by recent developments in remote sensing and computational sciences. Most applications rely on one or more numerical models that transform image data in desired thematic information.
Building performant models requires high-quality training data (also called reference or ground truth data). Acquiring this data unfortunately takes up a lot of time and effort, in part due to the absence of a standardized approach. Less attention has been paid so far to how we can generate training data in a smarter way.
GENLIB is a research project that addresses this issue. We essentially want to help make mapping of urban environments with optical remote sensing data more efficient and accessible, including for non-expert users. Particular focus is put on automation of training data acquisition, management and optimization.
At the core of this project lies the concept of a spectral library...
A spectral library?
A spectral library is a collection of labelled spectra that is usually accompanied by at least some metadata, e.g. on how, when and where these spectra were acquired.
Spectra or spectral profiles describe the amount of sunlight that is reflected/absorbed by different surfaces. Reflectance can vary over wavelengths depending on the (bio)physical or chemical properties of the surface being measured, resulting in unique signatures that allow us to identify those surfaces (and sometimes even their condition).
If a surface type is pure, meaning it consists of only 1 material, we call it an endmember. Spectral libraries typically contain only pure endmember spectra labelled with their corresponding cover type. Labelled endmember spectra are essential for many urban mapping applications.
Endmember spectra can be measured with laboratory/field spectrometers (near or on the ground) or sampled from remote sensing imagery (from the sky). While ground measurements are typically more accurate, they are also more expensive to acquire.
High-resolution imagery, i.e. having both small pixels (<5m) and many narrow bands, is most suited for building spectral libraries. For now, such imagery is acquired with airborne sensors, although in the near future this may also become possible with spaceborne sensors ...
How we mostly use spectral libraries
(and why it's a problem ...)
This is what happens now: we all work on our little image
spend a lot of time and effort to get suitable
finally we get that library
make something with it
and then stop using the library altogether.
WHAT A WASTE!!!
Obviously this is very inefficient. On the one hand we are doing the same work over and over again, even though it can often be avoided. On the other hand, there are those who don't have the skills or data access to make those libraries, but could definitely use them for their applications.
What we propose
In GENLIB we put forward the idea of a Generic Urban Library (GUL) to address the issues mentioned above. A GUL is a collection of multi-source urban spectral libraries, each provided with ample metadata and structured in a common format.
The GUL is equiped with a toolset that facilitates automatic extraction and labeling of endmember spectra from remote sensing imagery. It also allows GUL spectra to be used dynamically for various applications on multiple sites, using different sensors. The toolset optimizes the GUL for the task at hand.
The GUL also features an online platform with an interface that allows users to share, review or query spectra. Crowdsourcing improves the quantity and quality of the GUL.
This is the vision that we want to contribute to.
To sum up ...
We all bring our images
use GUL tools to extract and
label endmember spectra
get a bunch of spectral libraries
consolidate those libraries in an online GUL environment
then use other GUL tools to optimize endmember
spectra for specific tasks
to support your urban mapping application.
The main hypothesis of GENLIB is that a Generic Urban Library, once it becomes sufficiently representative, can be used to perform accurate mapping with any optical sensor, on any site and for any application, assuming that the used image data are reasonably suited for the task.
To test this hypothesis, we formulate the following operational objectives:
Define a metadata model that encompasses all the information needed for processing and querying.
Propose a hierarchical labeling scheme for urban cover types that is conform with the EAGLE framework.
Provide a toolset for dynamic use of the GUL, including expansion, management and optimization.
Build a first version of the GUL using a multi-site collection of urban hyperspectral imagery.
Deliver a proof of concept by performing a series of mapping experiments with the GUL.
Want to know more?
Like what you see? Got some interesting questions or remarks? Or just want to compliment our nice little project?
You are more than welcome to contact us via the GENLIB project page on Research Gate. If you follow us there, you will be kept up to speed with our latest activities and publications.
Also, we are planning to attend at least the following events:
Meet us there!
We are a Belgian-German consortium comprising 3 research institutes with experience in the fields of urban remote sensing, image processing and spatial analysis.
Cartography and GIS research group
Division of Forest, Nature and Landscape
Earth Observation Center