What is Change Detection and how I can perform such analysis with open source tools?

Question

From wikipedia page:

Change detection for GIS (geographical information systems) is a process that measures how the attributes of a particular area have changed between two or more time periods. Change detection often involves comparing aerial photographs or satellite imagery of the area taken at different times. The process is most frequently associated with environmental monitoring, natural resource management, or measuring urban development

How is the comparison done? With what tools? I feel that the description is not complete. Or something is missing.

Where or in which books can I find more information about Change Detection?

What tools should I use to perform such an analysis using the data in a shapefile? (only open-source please)

EDIT

Some papers on change detection (theory and techniques)

Change detection techniques (D. LU, E. BRONDI, ZIO and E. MORAN, 2004, pdf)

(I'll add more in the future as If I discover more notable papers)

Answer 1

Change detection is a common operation/module in remote sensing packages like ENVI or Orfeo toolbox. It usually involves raster data (satellite images for example).

How is the comparison done? With what tools? I feel that the description is not complete. Or something is missing.

Change detection is done by comparing two raster images that were taken at different times but which cover the same area. As the images cover the same area, the images overlay each other. Imagine two grids stacked on top of each other.

It is then a matter of comparing whether the value of a pixel in the new raster is the same as the value of the pixel in the old raster. Pixels that have changed are then marked. The output is usually a raster that covers the same extents as the two images with the changed areas highlighted. It's a simplification of course but you get the idea :)

Where or in which books can I find more information about Change Detection?

You can start with these documents

• Review Article Digital change detection techniques using remotely-sensed data
• THE STATE OF CHANGE DETECTION IN GIS
• How Change Detection Works

What tools should I use to perform such an analysis using the data in a shapefile? (only open-source please)

You can try out Opticks. It has a change detection plugin.

Answer 2

Check out DTclassifier here which you can use with QGIS.

DTclassifier provides simple streamlined interface for raster classification and change detection using decision trees.

Plugin features:

• integrated approach — perform all operations including training data collection, tree-building and classification in QGIS
• first example of using computer vision library OpenCV in QGIS
• use of non-parametric classification algorithm — decision trees.

You can find a tutorial here.

Beside this you can glance at this post here, Entropy change detection

I hope it helps you...

Answer 3

I don't think you will find many tools for change detection on vector data (like shapefiles) because its a trivial problem - just walk the points, and tell me if they are the same.

Change detection is more typical for raster images (e.g. SAR images, or visual/IR images), where the problem is detecting what has changed from one satellite pass to the next, or from one aircraft overflight to the next, or "before and after" on a site that has experienced natural disaster.

For raster images, one open source toolkit option is Orfeo Toolbox.

Answer 4

Change detection

Change detection, in the Remote Sensing discipline, is the analytical process that aims to detect changes -- over time and space -- of the land cover or/and land use.

PCA as a change detection technique

Among the most common and successful change detection practices, is the application of Principal Components Analysis (PCA) on bi- or multi-temporal multi-dimensional data (Lu et al., 2003).

What is PCA?

Principal Components Analysis (PCA) is a multi-dimensional linear transformation algorithm. It reconstructs a multivariate data set in a way that the first variables, called principal components (PCs), contain most of the original data variance. Thus, PCA provides the potential to describe or represent reliably a multi-dimensional data set by using fewer dimensions than the ones that compose the initial data set (Jolliffe, 2002).

How does it work?

PCA redirects the highest variances of the original data set, which mainly resemble unchanged landscape characteristics, in the first components. It is the user's responsibility to then extract changes by means of advanced digital image processing operations, i.e. image (segmentation and) classification.

PCA-based change detection using (G)FOSS

PCA is implemented in GRASS-GIS (i.pca module), R (princomp() and prcomp() functions), OrfeoToolbox, SAGA-GIS and probably more (Free &) Open Source Applications.

An example in-depth work, from which most of the above text has been extracted, demonstrates how to map burned areas -- which is essentially a change detection analysis -- based on PCA and GFOSS. Please, refer to this work for an extensive list of references upon the subject.

On the use of GRASS-GIS and R to perform PCA, there is a dedicated GRASS-wiki page titled Principal Components Analysis.

References

Jolliffe, I. T. (2002). Principal Component Analysis. Springer, 2nd edition. 28 illustrations.

Lu, D., Mausel, P., Brondizio, E., and Moran, E. (2003). Change detection techniques. International Journal of Remote Sensing, 25(12):2365.