Remote Sensing II, autumn 2005 - spring 2006
In case of computer problems, user licenses, etc. contact Juho Kolari, jkolari@cc.hut.fi, M315.
Part 1: Download IMAGE2000
Download IMAGE2000 Landsat ETM-images from Joint Research Center,
and write them to DVD. Use computer mato.foto.hut.fi (MS Windows) which is in room M277. Acquire user license
for computer from Juho Kolari. You have also to acquire license for JRC-imageserver. Images, both multispectral
and panchromatic, which should be loaded are those available from following list:
- path 185, rows 15-18
- path 186, rows 15-18
- path 187, rows 14-18
- path 188, rows 13-18
- path 189, rows 12-18
- path 190, rows 11-18
- path 191, rows 11-18
- path 192, rows 11-18
- path 193, rows 11-14
- path 194, rows 11-13
- path 195, rows 11-12
- path 196, rows 11-12
Images should be downloaded to directory "Shared documents: Maa57252" or "D:Maa57252", they should be same.
After downloading suitable number of images, write them to DVD. Contact Juho kolari if problems.
Part 2: Processing of Landsat ETM-image mosaics
Mosaic ETM-images to regional mosaics according to regional areas (maakunnat), one area per student.
this will be performed after student computers are working and new PCI Geomatica installed, hopefully
within next month. Aim is to get radiometrically corrected reflectances from images. Steps:
- Cloud and shadow masking, manually or by clustering
- Atmospheric correction using PCI
- Topographic correction, is it needed?
- Jenni Lampila's Master thesis, 2000
- Törmä, Härmä: Topograhic correction of Landsat ETM-image sin Finnish Lapland, IGARSS 2003
- Mosaicing
Instructions and helpful information:
Digital map data (send me Your introductions):
- Topographic database: Valtteri
- Soil and rock: Maija-Liisa
- Digital Elevation Model: Matthieu
- Digiroad and PerusCD: Jussi
Whan to do next and some more instructions:
- Copy downloaded images to /home/courses/maa57252/landsat/
- Choose natural province and communicate it to Markus
- Select images which cover chosen province
- Transfer them to PCI-format
- Go to Focus-window
- File -> Utility -> Import to PCIDSK
- select BSQ-image file, give pci-filename, then go on
- define raw
- give parameters: header bytes 0, band interleaved, other info from image files
- Import
- Process images (atmospheric correction, topograpic correction?) and mosaic
- After mosaicking, cut areas which do not belong to chosen province away
Part 3: Interpretation of Landsat ETM-image(s)
Interprete land cover/use of some city with surroundings (e.g. Tampere) using Landsat ETM-images
and available GIS-data. Classes are Corine Land Cover 2000 Level 2/3 classes and they are
combination of land cover and use:
- 1. Artificial surfaces
- 1.1 Urban fabric
- 1.2 Industrial, commercial and transport
- 1.3 Mine, dump and construction sites
- 1.4 Artificial, non-agricultural vegetated areas
- 2. Agricultural areas
- 2.1 Arable land
- 2.3 Pastures
- 3. Forests and seminatural areas
- 3.1 Forests
- 3.1.1 Broad-leaved forest
- 3.1.2 Coniferous forest
- 3.1.3 Mixed forest
- 3.2 Shrub and/or herbaceous vegetation associations
- 3.2.1 Natural grassland
- 3.2.2 Moors and heathland
- 3.2.4 Transitional woodland/shrub
- 3.3 Open spaces with little or no vegetation
- 3.3.1 Beaches, dunes, and sand plains
- 3.3.2 Bare rock
- 3.3.3 Sparsely vegetated areas
- 4. Wetlands
- 4.1. Inland wetlands
- 4.2 Coastal wetlands
- 5. Water bodies
- 5.1 Inland water
- 5.2 Marine waters
More information about CLC2000 can be found from www.ymparisto.fi,
class descriptions (part 1
and part 2) from EEA.
Ground truth for classes can be acquired from map-servers of cities (aerial images available in many cases),
paikkatietolainaamo,
luonaispaikka, paper maps, etc.
Rough workflow:
- Select area (city with surroundings)
- Cut area from image mosaic if available, or perform image corrections to original image(s)
- Institute of Cartography and Geoinformatics has Digiroad, soil and rock databases, topographic database, some classes can be acquired from these
- Acquire ground truth
- Feature extraction and selection
- Supervised classification using Maximum Likelihood
- Unsupervised classification using K-means
- Accuracy assessment, compare to CLC2000
- Report results in a form of scientific article
Part 4: Change detection
Change detection of interpreted area, using old Landsat Thematic Mapper (about year 1990) and
MultiSpectral Scanner images (about year 1975). Test simple change detection methods and compare.
Old images can be found from Global Land Cover Facility.
Report results in a form of scientific article.
Part 5: Rule-based classification
Design rule-based classification using Dempster-Shafer method and experiences of previous parts with
pen and paper. Construct hierarchy of classification using previously presented CLC2000-classes.
Form rules for each set of classes to be classified, so that the result would be good. Material for classification
is Landsat ETM-images, digital elevation model, soil map, road database and topographic database (maastotietokanta).
Present methodology which combines the evidences of individual rules.
Useful literature:
- A.Srinivasan, J.A.Richards:
Knowledge-based techniques for multi-source classification.
International Journal of Remote Sensing, Vol. 11, No. 3, pp. 505-525, 1990
- G.G.Wilkinson, J.Megier:
Evidential reasoning in a pixel classification hierarchy - a potential method for integrating image classifiers and expert system rules based on geographic context.
International Journal of Remote Sensing, Vol. 11, No. 10, pp. 1963-1968, 1990
- G.Shafer, R.Logan:
Implementing Dempster's rule for hierarchial evidence.
Artificial Intelligence, Vol. 33, No. 3, pp. 271-298, 1987
- G.Shafer: A Mathematical Theory of Evidence. Princeton University Press, 1976