If your work is about invasive species, you may be interested in the PhD thesis of C.M. Joshi, who detected invasive Banmara (Chromolaena Odorata) in forests in Nepal using landsat images. You can find his thesis here:

http://www.itc.nl/library/Papers_2006/phd/joshi.pdf

Dear Vijayasekaran, 

I hope you do understand that every entity on earth reflects distinctly, because of which we obtain a different DN value in the satellite images. So, do different species of vegetation. Also, the spatial resolution matters, and it will not be of use if you want to do species discrimination using a Landsat image. Also, I do not support use of simple NDVI or vegetation indices for species discrimination, neither the common supervised or unsupervised classification techniques. That won't help! But you may try high resolution optical images and object based classification techniques, which utilizes many other features (texture, shape etc.) along with the reflectance values and indices. You can find several papers related to that. 

You can also try spectral unmixing techniques which are specifically used for kind of analysis you want to do, although it is not very simple and researchers often failed to do it correctly.  

Hope that helps!

Dear Chitrini Mozumder

Thanks for the answer. I understand what you have explained in detail. But I need to track the historical dynamics of a plant species like from 1970-2010. So it is better to go with Landsat image isn't it. The idea behind my question is to classify the plant species using medium resolution satellite image like Landsat combined with the NDVI transformed image. I want to know that including NDVI with the satellite image have any advantage or improve the accuracy of classification

Thanks 

Dear Vijayasekaran Duraisamy,

I used Landsat data (2, 4, 5 and 7, over 10 years) to map a French mountain chain called "Les Pyrénées" and we managed to map specific vegetation classes (moors for example) with a correct accuracy (overall accuracy of 70-80%).

In that study, we used multiple temporal image about the year (4 to 9) and different spectral bands available on Landsat (B, G, R, NIR and MIR).

We tried to add NDVI and other vegetation indexes but they did not managed to improve classification results. In fact, vegetation index are used if you want to separate vegetation/non-vegetation classes. If you want to unmixed vegetation classes, try to use all the spectral bands or textures like Christini said. Elevation models could be an answer too if you are in mountain areas.

If your work is about invasive species, you may be interested in the PhD thesis of C.M. Joshi, who detected invasive Banmara (Chromolaena Odorata) in forests in Nepal using landsat images. You can find his thesis here:

http://www.itc.nl/library/Papers_2006/phd/joshi.pdf

Dear Researchers,

Thanks for your valuable suggestions. 

I appreciate the idea given by  all of you. I try to execute the same. I planned to include textural measures like GLCM and including NDVI for my proposed classification.

Thanks

VJ

Hello Vijay,

I agree with the suggestions given.  The NDVI by itself (from a single date) might not be useful, but if there is a contrast in the phenological cycle of your species of interest with others (and also that it has a good coverage given the relatively low resolution of Landsat imagery), then you can use time series of NDVI and see if the patterns in NDVI variation help to distinguish it from other species-land cover.  Hope this is useful.

If you have the spectral signature of the species that are you working with, the classification process would be easier. However, the spatial and spectral resolution of the remote sensing data affects the results, because of the bands that you are using for the image classification. If you are working with Landsat data, try to use the 4-5-3 band combination, which is normally used for landuse and landcover analysis.  

https://www.researchgate.net/publication/309589918_Driving_force_for_forest_fragmentation_explored_by_land_use_change_in_Song_watershed_India

Article Driving force for forest fragmentation explored by land use ...