Cairo - Exercises using Landsat data - Continued
|  | False-colour combination image of Cairo using bands 7,4,2 | |
False-Colour Combination
To increase the interpretability of satellite images, false-colour images are frequently used.
In most cases, a false-colour image uses at least one infrared channel. The infrared range is very useful for interpreting the Earth's surface, because it consists of reflected and emitted energy.
Infrared light is not visible to the human eye but reveals a lot of information. In particular, plants reflect much more energy in the near infrared than in the visible range of the electromagnetic spectrum. Even the health status of a plant can be ascertained from the intensity.
Open the LEOWorks programme. If you have not downloaded the images of Cairo yet, do so now.
Choose File>Open. A dialogue box will pop up. Choose the folder Cairo and select the first image Cairo_Landsat_2000_Band_2.tif. Open Cairo_Landsat_2000_Band_4.tif and Cairo_Landsat_2000_Band_7.tif, too.
Choose Image>Combine from...>Red Green Blue. A pop-up menu will open. Select image Cairo_Landsat_2000_Band_7.tif for Red, Cairo_Landsat_2000_Band_4.tif for Green and Cairo_Landsat_2000_Band_2.tif for Blue and click OK. Improve the raw data in the same way you did for the True-Colour Combination exercise.
This new image is a false-colour combination of three greyscale pictures.
| | | False-colour combination image of Cairo using bands 4,2,1 |
Describe the image and try to divide the image features into 5 classes - parks and farming areas, water, desert and plain soil, densely built-up areas, loosely built-up areas.
How did the colours of the classes change between the true- and the false-colour images?
Can the classes be identified in both images?
Which image would you prefer to differentiate between the satellite image's features?
Try another combination.
Choose Image>Combine from...>Red Green Blue. A pop-up menu will open. Select image Cairo_Landsat_2000_Band_4.tif for Red, Cairo_Landsat_2000_Band_2.tif for Green, and Cairo_Landsat_2000_Band_1.tif for Blue, and click OK. Improve the raw data the same way as described in the exercise True-Colour Combination.
This new image is a false-colour combination of three greyscale pictures.
Describe the image and try to divide the image features into 5 classes - parks and farming areas, water, desert and plain soil, densely built-up areas, loosely built-up areas, and compare them with the true-colour image.
Describe the differences between the two false-colour images.
Select the 5 classes in both false-colour images and compare the colours used in the images.
Combine a false-colour image by using channel 7 for Red, channel 1 for Green and channel 2 for Blue.
What can you say about the usefulness of the image?
Now try a combination of your choice and note the variations.
| | Parallelepiped classification of Cairo | |
Multispectral Image Classification
In exercise 2 you separated objects and elements of Cairo's surface into classes. To interpret the features of a satellite image, a visualised classification is very useful. The classified image is, with some basic revisions, similar to a thematic map. The main classes of Cairo's surface are densely built-up areas, loosely built-up areas, forest, agricultural land, water, desert and bare soil.
There are different classification tools, based on different image processing methods. Which method is used depends on what information is required.
Learn more about image classification with the LEOWorks Tutorial.
Parallelepiped classification
The parallelepiped classification is a simple supervised classification method. It is based on the spectral ranges of the different land use classes within the different bands. It uses training fields representing the different land use classes. Each land use class has its specific spectral fingerprint. For each type of land cover to be classified, at least one training field has to be defined. LEOWorks evaluates all these training fields and allocates each image element (pixel) to one of the given land use classes.
Open the LEOWorks programme. If you have not downloaded the images of Cairo yet, do so now.
Choose File>Open. A dialogue box will pop up. Choose the folder Cairo and select Cairo_Landsat_2000_Band_1.tif.
Cairo_Landsat_2000_Band_2.tif, Cairo_Landsat_2000_Band_3.tif,
Cairo_Landsat_2000_Band_4.tif, Cairo_Landsat_2000_Band_5.tif, and
Cairo_Landsat_2000_Band_7.tif. Improve the raw data the same way you did for the True-Colour Combination exercise.
Select (activate) image Cairo_Landsat_2000_Band_7.tif.
Choose Multivariate Analysis>Supervised Classification>Select Training Fields.
Select Draw Polygon from the toolbar.
Note: The more training fields you are able to select for a class, the more accurate the result will be.
Start with water surfaces and draw a polygon inside the River Nile. Name the class Water. Find another part to draw a second polygon inside a water body and name the class Water, too, and so on. When you are done with the water, draw a polygon inside an area covered by dense desert and name the class Desert. Do the same with all 7 pre-selected main classes.
Choose Multivariate Analysis>Supervised Classification>Parallelepiped and select all images.
Write down the 8 different colours and add the class names. It might be useful to open the image Cairo_Landsat_2000_Band_321.tif for comparison.
Choose Image>Add Legend and check the combination of colours and class names.
Save the classified image as cairo_class_parallel (TIF) in the folder Cairo.
Discuss the accuracy of the supervised parallelepiped classification.
Which sources of errors could be involved?
| | | Maximum likelihood classification of Cairo |
Maximum Likelihood
The maximum likelihood classification is another supervised classification method. It is based on sophisticated statistical methods. It also uses training fields representing the different land use classes. Each land use class has its specific spectral fingerprint. For each type of land cover to be classified, at least one training field has to be defined. LEOWorks evaluates all of these training fields and allocates each image element (pixel) to one of the given land use classes.
Open the LEOWorks programme. If you have not downloaded the images of Cairo yet, do so now.
Choose File>Open. A dialogue box will pop up. Choose the folder Cairo and select the first image Cairo_Landsat_2000_Band_1.tif. Open Cairo_Landsat_2000_Band_2.tif, Cairo_Landsat_2000_Band_3.tif, Cairo_Landsat_2000_Band_4.tif, Cairo_Landsat_2000_Band_5.tif and Cairo_Landsat_2000_Band_7.tif, too. Improve the raw data the same way you did in the True-Colour Combination exercise.
Use the same training fields as in the Parallelepiped Classification exercise.
Choose Multivariate Analysis>Supervised Classification>Maximum Likelihood and select all images (except cairo_class_parallel.tif if still open).
Set the Threshold Value to 5%.
Write down the 8 different colours and add the class names. It might be useful to open image Cairo_Landsat_2000_Band_321.tif for comparison.
Choose Image>Add Legend and check the combination of colours and class names.
Discuss the accuracy of the supervised maximum likelihood classification. Which sources of errors could be involved?
Compare parallelepiped and maximum likelihood classifications and discuss how they agree and how they differ.
Try other threshold values, for example 50% and 75%, and compare the classified images.
| 
