Skip to content

Multitemporal Remote Sensing Feature Mining for Machine Learning Applications

License

Notifications You must be signed in to change notification settings

shepherdmeng/ROIseries

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ROIseries

data/info_graphics/ROIseries_video_thumbnail.png

Purpose

ROIseries helps to discover usefull features in multitemporal remote sensing data for machine learning applications.

Installation

  1. Either A: clone the ROIseries repository or B: download the latest release (source code and data.zip)

  2. If A: Make sure that the data in ROIseries/data/ is pulled down using git LFS.

    If B: After unpacking the source code and the data.zip, replace the data folder within the source code with the unpacked data.zip.

    Note: The data is only necessary for the examples, ROIseries will work without the data

  3. Add the ROIseries folder to your IDL Paths in the IDL Workbench.

  4. DONE :)

  5. To get started and/or test if everything works, try out the examples in: ROIseries/docs/examples/

Tutorials

Have a look at the docs/examples folder. It contains tutorials and Jupyter Notebooks demonstrating the tutorials e. g.:

Structure

  • data: Contains data used in the examples and for testing
  • docs: Contains examples and other documentation
  • src: Contains the ROIseries classes + routines they access.
    • book_keeper: Tabular input/ouput
    • cookie_cutter: Extract spatiotemporal image-objects from a series of rasters
    • features_sommelier: Cross-validation using machine learning to assess feature usefulness
    • glcm_features: Calculate the gray level co-occurrence matrix + features
    • spatial_mixer: Quantify the numeric distribution in the spatial dimension while removing it (ROIseries_3D->ROIseries_1D)
    • spectral_indexer: Extract or mix raster bands from a mulitspectral raster to remove the spectral dimension.
    • temporal_blender: Arithmetically combine different ROIseries objects.
    • temporal_filter: Apply filter along the temporal dimension.
    • sub_routines: Collection of usefull routines used throughout ROIseries.
    • visual_tukey_: Visualization of ROIseries_3D and ROIseries_1D objects.
  • tests: Contains tests.

Background

ROIseries is developed at RLP AgroScience GmbH (http://www.agroscience.de/) to serve the requirements of various projects in the field of land cover classification. It is applied for example in the NATFLO project (http://www.natflo.de/). It is also the topic of a master thesis written within the UNIGIS distance learning program (http://salzburg.unigis.net/).

Requirements

ROIseries is mainly written in the Interactive Data Language (IDL) and requires an IDL development license. The machine learning feature cross validation is written in Python + additional Python packages.

Contributing

Everybody with any level of expertise in IDL and Python is most cordially welcome to contribute to the library with questions, constructive criticism and code contributions. For more information about how to contribute please read CONTRIBUTING.

License

GNU AFFERO GENERAL PUBLIC LICENSE Version 3, 19 November 2007 (cp. LICENSE)

About

Multitemporal Remote Sensing Feature Mining for Machine Learning Applications

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • IDL 80.0%
  • Python 20.0%