Change Detection Software for SAR and Optical Images

Behnaz Pirzamanbein, Allan Aasbjerg Nielsen

AbstractChange detection is one of the important tasks in Earth observation and monitoring. Analysing the changes in synthetic aperture radar (SAR) and multi-spectral optical images through different time points guide us in discovery of significant environmental events, and in managing forest and agricultural lands. In this paper we present two standalone software, WISHARTChange and MADChange based on two well-known change detection methods, the omnibus test and the iteratively re-weighted multivariate alteration detection (IR-MAD), respectively. The omnibus test method deals with a time series of SAR data and computes a sequence of test statistics for covariance matrices. IR-MAD deals with multi-spectral optical images and computes the changes between two time points.

Given the availability of Earth Observation (EO) data from different sources, the standalone software is developed to handle different formats of images such as Georeferenced Tagged Image File Format (GeoTIFF), and ENVI binary image coupled with a header file. Moreover, to overcome the big EO data challenges, the software computes the changes in two different processing modalities: 1) reads the whole images into local memory, and 2) treats the data line by line. In addition, due to independent computations of test statistics in WISHARTChange software, we adopt a parallel computing scheme.

The software is published on "" in two versions; GUI app and command-line executable. The implementation is done in MATLAB, however the software runs "without" MATLAB being installed on the local computer. The user only needs to install the MATLAB Runtime which is provided with the software and it contains all the functions needed to run.

In both WISHARTChange and MADChange users are required to provide SAR or multi-spectral optical images and can choose processing modality. More specifically, in WISHARTChange, equivalent number of looks, polarization type and names, and time points name must be specified. In addition, there is an option to select and compute the changes in a region of interest (ROI) by providing a binary mask or by choosing the ROI interactively from a provided map. In MADChange, the name of multi spectral bands is required and a threshold as a criterion to stop the iteration can be specified. Furthermore, there are two pre-processing scheme options: 1) masking the strongest changes and 2) excluding low values related to dark regions which can be used.

Each standalone software outputs maps of identified changes in multiple formats. The WISHARTChange software outputs a table containing average no-change probabilities and a figure containing three maps: first change, last change and frequency of the change. The MADChange software outputs a figure showing the probability of no-change and the canonical correlation convergence plot. Moreover, the IR-MAD variates are saved for further analysis as an image with same number of bands and same format as the provided images.

The maps of detected changes provide a better insight into analysing and monitoring of spatio-temporal dynamics for the area of the study which assists environmental managers and policy makers in decision making.

(Footnote: This work is funded by, DataBio, the European Unions Horizon 2020 research and innovation programme under grant agreement No 732064.)
TypeConference paper [Abstract]
ConferenceESA Living Planet Symposium
Year2019    Month May
AddressMilan, Italy
BibTeX data [bibtex]
IMM Group(s)Image Analysis & Computer Graphics