I have done LULC classification using LANDSAT images for my study area which includes a part of coastal seashore. The classification was done using 2 methods: the conventional supervised MLC method and spectral indices based classification. In both the cases it is found that intermixing of built-up, barren and coastal shore classes. Coastal shore appeared to be built-up class and barren to be again built-up class. Though it is claimed by many of the researchers that spectral indices based classification gave good overall accuracy and Kappa coefficient, the problem mentioned above still persist. It is observed in my current work that overall accuracy and kappa coefficient were enhanced to an extent. But, the problem of intermixing of pixels is more prevalent especially in coastal area.
Please let me know if you have any idea on how to tackle this issue.
Thank You.
Try the following as a starting point:
https://pdfs.semanticscholar.org/63b9/7ce83281e012da96d7be39d84a1739fae67c.pdf
Suggested further readings:
Article Potential of Landsat-8 and Sentinel-2A Composite for Land Us...
https://www.mdpi.com/2072-4292/1/3/330/pdf
It is a common mistake for users, I suggest you choose your training area more carefully and use Support Vector Machine algorithm in ENVI environment preferably version of 5.5 as you can modify the classification results on your own and re-class the mistakenly classified pixels. If you do not have Envi 5.5 then you may need to transport your classified map in to arc or Erdas for correction of the wrongly classified pixels. In Erdas you can use Recode option for doing so.
Article Modelling the forest transition in Central Western Ghats, India
Dear when you perform classification, it is already give you the full thematic map and add to that thematic map for each class
regards
Hi Vineela Nandam,
You could use eCogntion Developer 9.5 or eCognition Essential to applying the Image segmentation and the Object-Based Image Classification, then you will get the best classification for your study area. Good luck. Regards.
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Atmosphere