IMPRS Summerschool 2018

IMPRS 2018 MPI BGC Jena, 3.9.2018

PD Dr. S. Hese (Friedrich-Schiller-University Jena, Department for Earth Observation,

Introduction to image processing for Earth observation data analysis: – Landcover Classification and Multispectral Data Analysis using the ENVI/IDL Image Processing Environment

Data: Landsat TM5 subset (LS-TM5-02071993-SUBSET-UTM32.envi.hdr), 30 m spatial resolution, UTM 32, Datum:WGS84, BSQ file structure.


ENVI 5.4 Hands-On Exercises:

  1. Software Environment Setup:
    1. ENVI working directory setup
    2. ENVI .hdr & external format import ()
    3. Linking displays within ENVI
    4. Window Management in ENVI Classic
  2. Subsetting
    1. Data spectral sub-setting (deleting channel 6)
    2. Spatial sub-setting using ROI Tools
  3. Header Data Analysis
    1. spectral properties,
    2. geocoding information,
    3. header information,
    4. band statistics & histograms,
    5. band scatterplots
    6. BSQ /BIL Setup
  4. Data Preprocessing
    1. Orthocorrection & ATCOR (demo in PCI EASI/PACE)
    2. Contrast stretching / histogram modifications on display and with LUT
    3. Filtering with spatial filters, low pass, high pass, custom filter functions)
  5. Texture filtering with Co-Occurrence masks
  6. “Band Math” tool syntax (IDL) and NDVI calculation into a new floating point image file
  7. Landcover classification
    1. Preparation and design of ROIs for supervised classification,
  8. ROI management
  9. Signature separabilit
  10. Scatterplot visualisation
  11. n-dimensional visualization
    1. Maximum likelihood classification with ROIs
    2. Calculation of area statistics from thematic classification results

Post-classification analysis – confusion matrix using ROI ground truth