Did you mean when you make a classification can not distinguish between buildup area and rocks?

Yes, in case of pixel based supervised classification.

Hello,

For the supervised classification, did you specify a number of classes before starting the classification? Also, did you work with the signature of the classa used? Which software do you work with? I await your answer:)

Yup I am working with spectral signatures of different features, other details are as follow:

- Number of classes = 8 (Forest, Scrubs, Builtup, Barren, Agriculture Crops, Grasses & Hurbs, Waterbody, Snow & Ice

- I am using ERDAS Imagine

- Satellite Image: Landsat TM 30 meter (Surface Radiance)

Yes, but sometimes it not perfect or easy to define a class...mainly in transition forest fallow....or degraded forest and fallow are a in tropical région...

Second question, tes...

I am using Erdas engine

You can not classify arid or bare lands and built-up areas simply by providing clues that it is a non-linear problem. Therefore you can train the system with some sets of classified entries and use your test data, you can get the desired result.

With similar studies, he used the Impervious Surface Fraction (ISF) to estimate built-up areas. it has shown that deriving ISF from Landsat 7 ETM + is quite sophisticated, it can be applied to OLI data. However, built-up areas can be either impervious or permeable around cities. it depends on the purpose of your study.

In addition, the analysis of time series and the determination of the appropriate time make it possible to solve this type of problem. In addition, you can do this based on the thermal band as it is useful for this type of study.

Cordialy

Jad Tahouri

Thank you Jad

" Therefore you can train the system with some sets of classified entries and use your test data, you can get the desired result."

Can you please explain a bit more?

Instead of applying supervised classification in ERDAS imagine I would like to suggest you, please try GEOBIA or RF classification. You can easily run few lines code in Google Earth Engine to retrieve impressive results. In case of any further assistance, please free to let me know.

Thank you Dr. Hammad for good suggestion. Let ma explore GEOBIA and then will return to you for any technical query.

use pan sharping to get more reliability on spatially aspect. Build up areas mostly lies within 15m domain while barren areas may more than 15m. For more convenient if you consider for RS in urbanization use my paper on lahore

https://www.researchgate.net/publication/303842478_Urban_change_detection_of_Lahore_Pakistan_using_the_Thematic_Mapper_Images_of_Landsat_since_1992-2010

Nasir Hameed,

Thank you for good information.

Usman Saleem

Actually I am using landsat surface reflectance product which is not available for panchromatic band and I think its not appropriate to merge reflectance values of multispectral image with DN values of panchromatic band. 2ndly, panchromatic band is not available for landsat5 TM.