Requesting guidance on land use land cover classification on Landsat imagery using Python

[nishadh]

Hi,

Land use land cover classification for India for the year 1985-2005 based on Landsat satellite imagery is available in public domain as geotiff imagery https://daac.ornl.gov/VEGETATION/guides/Decadal_LULC_India.html 

To get the latest year land use land cover statistics such as of 2015, requires need of classification on raw images. Python with its open source libraries such as rasterio, numpy and scipy is apt to carry out classification. There is a tutorial describing the various steps to address using Python https://www.machinalis.com/blog/python-for-geospatial-data-processing/

 

However assessing the classification accuracy and result comparable with above open data is no where near to expectations. Requesting guidance in this regard.

Regards,

Nishadh.K.A.

[Devdatta Tengshe]

Hi Nishad,

If you are going to classify Landsat Data, you should do a Supervised Classification. This means that you provide input points, whose Landuse you know.  This is an iterative method, where you need to provide inputs for the algorithm to give a better output.

The Link that you shared does Un-Supervised Classification, i.e. without any apriori knowledge, and hence the output quality will be lower.

I'll suggest you have a look at the Semi-Automatic Classification Plugin for QGIS. The website has excellent video tutorials which show you how you can achieve a decent LULC classification.

Regards,

Devdatta

--
Datameet is a community of Data Science enthusiasts in India. Know more about us by visiting http://datameet.org
---
You received this message because you are subscribed to the Google Groups "datameet" group.
To unsubscribe from this group and stop receiving emails from it, send an email to datameet+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[nishadh]

Hi Devdatta,

Thank you very much for the post, I will try with  Semi-Automatic Classification Plugin for QGIS. 

Regards,

Nishadh