|
|
Congo River Basin - Landsat data - Continued 
Unsupervised Classification
To improve planning and the use of available resources, an understanding of the environment is necessary. A relatively simple type of classification is the Unsupervised Classification. All pixels of an image are grouped into a specified number of classes based on the similarity of their greyscale values. Open the LEOWorks programme. If you have not downloaded the images of the Congo River Basin yet, do so now.
Choose File>Open. A dialogue box will pop up. Choose the folder Congo and select the images Congo_Landsat_Band_1.tif, Choose Multivariate Analysis>Unsupervised Classification and select all images. Type 5 for No. of Classes and 10 for No. of Iterations. An iteration is the repetition of a sequence of computer instructions a specified number of times or until a condition is met. Save the new image as congo_unsupervised (TIF) in the folder 'Congo'. Add a legend to the classified image. Choose Image>Add Legend and select the first class.
Try to find a name for each class by comparing the classified image with the true-colour Congo_Landsat_Band_321.tif image, and also with the result of your interpretation into the 5 proposed classes (dense forest, shrubs, clear water, muddy water, and clouds) in the Multispectral Image Combinations exercise.
Compare the chosen class names with those obtained by your classmates and explain your thoughts. All image information is re-sampled and grouped into the 5 classes. The assigned colours, however, are quite inoperative. Change the colours into more familiar ones. Choose Edit LookUp Table and select the first colour. Change the RGB values by using the slides. Try to find colours appropriate to the classes and save the image under the same name. Now try other variations, for example 4 or 6 classes with 10 iterations and 5 classes with 3 or 25 iterations.
Note and discuss the particular changes.
The classified image 'congo_unsupervised' is full of widespread colours and very tiny objects. In some parts of the image these objects are hardly distinguishable. Contours are useful for a better differentiation. Furthermore, contours allow magnification without the disadvantage that the elements become blurred. According to requirements, the contours are generated from different images. To specify the water surface, for example, the use of an infrared image is reasonable.
Open the LEOWorks programme. If you have not downloaded the images of the Congo River Basin yet, do so now.
Improve the image Congo_Landsat_Band_7.tif in the same way it was done in the True-Colour Combination exercise.
Choose Enhance>Filter>Find Edges>Sobel.... And then Image>Invert....The image will change into a line drawing. You can further enhance it using Enhance>Interactive Stretching.... Save the image as congo_contour7 (TIF) in the folder Congo.
Open the image congo_unsupervised.tif and convert it to RGB. Choose Image>Convert to...>Red Green Blue.
Choose Image>Transparent Overlay. Select the line drawing congo_contour7.tif as BackGround Image and the classified image congo_unsupervised.tif as ForeGround Image and use 25% for the opacity.
Which classes are framed by polygons? Explain this by thinking of the different wavelength reflection conditions. What can be said about the arrangement of the polygons and the different water surface colours? Compare the two line drawings (Contour 1 and Contour 2) here. They show quite different structures. Try to name the main elements of the images. Use the classified image congo_unsupervised.tif as a reference. Which of the two images is an infrared image? Explain why. Last update: 16 April 2013
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
