|
|
Bardia Region Overview - Image processing - Part 5  This excercise is divided into eight parts and requires the use of LEOWorks. Multispectral Image Classification The objective of this exercise is to learn about multispectral image classification techniques and Bardia in general. Digital image classification is a rather difficult task which requires good knowledge of the area treated in order to extract accurate information. Multispectral image classification is a useful and valuable method for the generation of thematic maps, and the highlighting of timely changes, like, for example, land cover. We will work with both, unsupervised and supervised classification. Unsupervised classification Unsupervised classification is a relatively crude method. In unsupervised classification, all pixels of an image are grouped into a specified number of classes based on the similarity of their greyscale values. If you have any problems with LEOWorks, consult the corresponding chapters in the Tutorial. Remember that there is also the Help button! In LEOWorks, open the following the images
Select type 10 for Nr. of Classes and 5 for Nr. of Iterations. The more iteration there is, the better similar classes are formed, but the higher the computing time needed. With Iteration 5 it might take a few minutes. As the result appears, apply Multivariate Analysis>Post classification filer>5x5. This will 'clean' the result. Save the new image as Bardia_2002_unsupervised (TIF) into the 'Bardia' folder. Add a legend to the classified image. Choose Image>Add Legend and select the first class. In the window 'Current Item', type in your choice of the land use compared to 'Bardia_Landsat_2002_Band_453.tif'. Try to give a name for each class by comparing the classified image with the false-colour image 'Bardia_Landsat_2002_Band_453.tif'. Compare it also with the result of your interpretation into the 5 proposed classes (forest, grass lands, agricultural lands, rivers and river beds) in the Multispectral Image Combinations exercise. There are 10 original classes. If different colours include the same land cover class, you can join classes by giving 2 or more classes (of the 10) the same colour.
1. Why are the original colours (class) of crop fields, grass lands and vegetation in the floodplain area similar? Can you explain?
Save the classification as 'Bardia_Landsat_2002_unsup.tif' in your 'Bardia' folder.
Last update: 16 April 2013
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Previous