extraction of built-up area from images

[imtest133]

Dear all

I'm new in remote sensing and I have problems when try to extract built-up area from an image or during image classification.

I am using landsat 7 ETM+ and TM images, when I use supervised or un-supervised methods for image classification, urban area (built-up area) are not recognizable from bare-soil.

I read several papers about built-up Indexes and tried some of them but there is no tangible improvement in results.

Please guide me as step by step process.

 

Regards

[intertronic]

hi, why do you use landsat 7 and not landsat 8? and with witch software?

 

 

here you can see the band combination for landsat 7 and 8

http://landsat.usgs.gov/L8_band_combos.php

http://web.pdx.edu/~emch/ip1/bandcombinations.html

 

so the band combination for TM is 1-4-7

in Landsat 7, the combination can be 7-6-4

http://www.harrisgeospatial.com/company/pressroom/blogs/tabid/836/artmid/2928/articleid/14305/the-many-band-combinations-of-landsat-8.aspx

 

you load every single band in ArcGIS and you do a combination

 

then do a classification (supervised or unsupervised)

 

Two major categories of image classification techniques include unsupervised (calculated by software) and supervised (human-guided) classification.

Unsupervised classification is where the outcomes (groupings of pixels with common characteristics) are based on the software analysis of an image without the user providing sample classes. The computer uses techniques to determine which pixels are related and groups them into classes. The user can specify which algorism the software will use and the desired number of output classes but otherwise does not aid in the classification process. However, the user must have knowledge of the area being classified when the groupings of pixels with common characteristics produced by the computer have to be related to actual features on the ground (such as wetlands, developed areas, coniferous forests, etc.).

Supervised classification is based on the idea that a user can select sample pixels in an image that are representative of specific classes and then direct the image processing software to use these training sites as references for the classification of all other pixels in the image. Training sites (also known as testing sets or input classes) are selected based on the knowledge of the user. The user also sets the bounds for how similar other pixels must be to group them together. These bounds are often set based on the spectral characteristics of the training area, plus or minus a certain increment (often based on "brightness" or strength of reflection in specific spectral bands). The user also designates the number of classes that the image is classified into. Many analysts use a combination of supervised and unsupervised classification processes to develop final output analysis and classified maps.

(source : https://articles.extension.org/pages/40214/whats-the-difference-between-a-supervised-and-unsupervised-image-classification)

[intertronic]

Edited June 20, 2016 by intertronic

[imtest133]

Thanks for useful information. I use landsat 7 because i need time series of images from 1985 to 2005. I tested Erdas imagine 2014 and Terrset for classification.

[Lurker]

you need a band combination that accentuate building area, thats all

[jalmiburung]

Thanks for useful information. I use landsat 7 because i need time series of images from 1985 to 2005. I tested Erdas imagine 2014 and Terrset for classification.

Hi,

 

You can use Building Index to extract urban area. There are many indices if you search the reference. You can download find one of them at http://www.mdpi.com/2072-4292/4/10/2957/pdf

[imtest133]

you need a band combination that accentuate building area, thats all

Dear lurker

I have tested various combinations but some of bareland recognized as urban area in both supervised and unsupervised methods.

How many bands is enough for image classification?

I used all 7 bands as layer stack. 

[Arhanghelul]

NDBI = Normalized Difference Built-up Index

search the formula for Landsat in Google

 

You can't escape the problem of mixt pixels between built-up and bare soil. These 2 elements have similar spectral signatures. You can only maximize the spectral differentiation by using derivates like NDVI, NDBI, NDWI, etc.... as aditional bands in Landsat scene classification.

 

There is some other supervised methos to better separate the elements, like Spectral Angle Mapper (SAM) , etc, etc.

search on internet and on ENVI help page, on sub-pixel classification methos.

 

Another method is to use OBIA > eCognition.

OBIA is based on image segmentation. Segmentation is grouping similar spectral pixels into objects. Then you classify these pixel groups based on their spectral signature, shape, area, texture, position into the image, etc.

OBIA is similar to visual image interpretation and classification

But some research articles say that OBIA is inferior to traditional per-pixel classification methods (supervised and unsupervised) in the case of medium and low spatial resolution images like Landsat, ASTER, SPOT 1-4, MODIS, etc.

OBIA works better on high-resolution images like WorldView-2&3, Quickbird, Pleiades, IKONOS, orthophotos, etc (<5 m spatial resolution). These images are much better for built-up area extraction compared to Landsat.

 

So, in your case, my friend, it depends on the accuracy of the classification that you want to obtain !

 

 

 

Edited June 21, 2016 by Arhanghelul

[imtest133]

NDBI = Normalized Difference Built-up Index

search the formula for Landsat in Google

 

You can't escape the problem of mixt pixels between built-up and bare soil. These 2 elements have similar spectral signatures. You can only maximize the spectral differentiation by using derivates like NDVI, NDBI, NDWI, etc.... as aditional bands in Landsat scene classification.

 

There is some other supervised methos to better separate the elements, like Spectral Angle Mapper (SAM) , etc, etc.

search on internet and on ENVI help page, on sub-pixel classification methos.

 

Another method is to use OBIA > eCognition.

OBIA is based on image segmentation. Segmentation is grouping similar spectral pixels into objects. Then you classify these pixel groups based on their spectral signature, shape, area, texture, position into the image, etc.

OBIA is similar to visual image interpretation and classification

But some research articles say that OBIA is inferior to traditional per-pixel classification methods (supervised and unsupervised) in the case of medium and low spatial resolution images like Landsat, ASTER, SPOT 1-4, MODIS, etc.

OBIA works better on high-resolution images like WorldView-2&3, Quickbird, Pleiades, IKONOS, orthophotos, etc (<5 m spatial resolution). These images are much better for built-up area extraction compared to Landsat.

 

So, in your case, my friend, it depends on the accuracy of the classification that you want to obtain !

 

 

 

Thanks for your comprehensive answer. I will try NDBI, NDVI, NDWI and such indices instead of raw bands of landsat in Suoervised and Unsupervised methods.

[deepgis]

NDWI and SAVI indices are very helpful in extraction of buildings. Choose the threshold values of NDWI to classify the buildings and after that use lower values of SAVI indices to remove the bare fields and bare rocks from previously classified building class.

[imtest133]

Dear deepgis

I used 3 PCA components of band, NDWI, NDVI, SAVI, 3 Tasseled Cap components all extracted from ETM+ band: 1,2,3,4,5,7 instead of raw bands. 

But i got only minor improvements in the results. Still in supervised and unsupervised barelands and urban area mixed together.

Please help me

[imtest133]

any more suggestion?

[deepgis]

Dear imtest33

 

I use reflectance data for this. I create NDWI using band calc (NIR-SWIR) / (NIR+SWIR) for the refectance data. For SAVI I use band calc (NIR-RED)*(1+0.5)/(NIR+RED+0.5).

[imtest133]

 

Dear imtest33

 

I use reflectance data for this. I create NDWI using band calc (NIR-SWIR) / (NIR+SWIR) for the refectance data. For SAVI I use band calc (NIR-RED)*(1+0.5)/(NIR+RED+0.5).

 

Dear deepgis

Thanks for your comment. you are right. I used reflectance data for calculating NDVI, NDWI and SAVI too and then combine them with 3 component of PCA of bands, as layer stack in Erdas imagine. 

[imtest133]

 

Dear imtest33

 

I use reflectance data for this. I create NDWI using band calc (NIR-SWIR) / (NIR+SWIR) for the refectance data. For SAVI I use band calc (NIR-RED)*(1+0.5)/(NIR+RED+0.5).

 

But only minor improvements in urban area.

[Arhanghelul]

Maybe there is a problem of content of moisture in the soil that makes buildings and soil very similar, so maybe you should select another Landsat scene from a dry season.

Have you tested with another Landsat scene for your area of interest ?

[imtest133]

Maybe there is a problem of content of moisture in the soil that makes buildings and soil very similar, so maybe you should select another Landsat scene from a dry season.

Have you tested with another Landsat scene for your area of interest ?

yes, I tried an image from 1985 and second from 2000. both from landsat. in both images I have problems in urban area and barelands.

Thanks 

[YumanFatima]

hey!!!! Can anyone tell me that what threshold value should I use if I have to extract built up area from an NDBI processed image?