Rajat,

You need to ask these questions.

1. Do you wish to consider each layer as one attribute value?

2. Why did you convert to ASCII values? The images have pixel intensities. So they must be numbers

I agree with above answer of Rajat. you need to have data in the form of records or columns. So i suggest you to copy the data into csv file first and then open it in weka.

@chandrasekhar,...I have clubbed together the layers into 1 single image,  now i have converted to ascii files, as after reading various document i came to this conclusion that WEKA accepts ascii files,

As .arff is also format of ascii , now when i tried copying the file which was in .TXT format into excel, it opens as data in one column which again doesn't solves the issue.

if you can help me, I will be much thankful

and when i have converted the Satellite image in ASCII format, it had few header information containing no. of rows and columns and no data value.

i have attached the screenshot which i get error while trying to input.

http://www.wikihow.com/Copy-Paste-Tab-Delimited-Text-Into-Excel

In that step 4, select 'space' instead of tab.

Also i have a suggestion for you. I have done a project on LISS (3 layers) satellite images in 2005 , for classification of vegetation type.

1.  Each layer has specific kind of information. So it's a bad idea to club all layers ( Depends on domain of the problem though). For example Heat signatures related classification problems need only infra red layer . Accuracy of the classification is actually disturbed if we club other layers.

2. Using matlab of Geomatica is best to work with images. Geomatica has tools to read, preprocess and cluster the satellite images( inbuilt). So there is no headache of transferring data to weka etc.

3. If you use matlab, you have facility to export the pixel values in 5 X matrix format directly to csv files with out the header info.

Now in this i have to extract Urban areas of Bangalore region from LANDSAT imagery now if i work on single band of Landsat Image, that is fine, i will figure that out which band is best for Carrying out Urban classification.

Secondly in MATLAB i had worked on Feature Extraction, I was unable to give full image as input therefore had to subset it, do i need to do in this also. 

ERDAS is the software i used in 2005 to do the Land cover detection of Allahabad. I have gone through your other posts also. Your Institute must  definitely having ERDAS, ARC GIS, ARC VIEW licenses. 

you first need to do Image pre processing like radiometric normalization and geometric rectification. Then follows the step of defining the area of interest and registering it with a previously surveyed map of bangalore region.The two images should be properly registered with the map achieving least possible RMS error.

There are two primary types of classification algorithm applied to remotely sensed data: unsupervised and supervised. Unsupervised classification algorithms (such as ISODATA), cluster data according to several user-defined statistical parameters in an iterative fashion until either some percentage of pixels remain unchanged or a maximum number of iterations has been performed. This method of classification is most useful when no previous knowledge or ground truth data of an area is available. However, the classes determined by the algorithm still require land cover identification by an experienced analyst.

I suggest you to use supervised clustering also to test.The output clusters may be assigned into one of the classes (urban, residential, barren land, grassland, forest, water, vegetation, mountain, riparian etc) with the help of the original image's spectral signature.

If you do not have domain knowledge of which band to use, you may use one layer at a time,  to compare the classification accuracies. You may next test with  combination of layers. Generally reflectance is criteria for urban region, water bodies and vegetation cover.

Hope this helps.

sir i have done all this earlier itself, but again my basic objective is to build a tool out open source software which does all that, as explained by you this full process is maual in nature

and suppose i have to classify 100 cities of india, again it will time consuming, so after many literature review i have found WEKA has a capability that it can automate classification, so i was trying to build a tool out of it.

I hope it is clear wat is my objective and data input is basic step which i am still not able to figure out. 

I'd say that the easiest approach is with R. It is free.

Something as simple as this.

read your Landsat 5 bands with

raster::stack(my_path_to_tiff)

foreign::write.arff(as.matrix(yourstack), file = 'mywekafile.arff')

A code for this:

library(raster)

library(foreign)

bands