At any instance can NDVI be used as a substitute to classification or to improve classification results?
hi, NDVI (= Normalized Difference Vegetation Index) in simple word, we use it to classified the vegetation, is based in the photosynthetically active vegetation to define the limited
between species, vegetation dynamics, total plant cover, biomass, and plant stress.
for me I use NDVI as a substitute for the classification
but if you want to know that NDVI improves or not it is necessary to use other software such as ENVI and make a comparison.
the difference between them is in utilization, for me is simple to manipulated with NDVI, for another is ENVI.
Salam
The relationship between NDVI and classification is what you make of it, or what the objectives of your classification are.
In other words, calculating NDVI for an image results in a continuous raster with values ranging from -1 to 1, where higher values indicate photosynthetically active vegetation (as Beloula Salima has mentioned above) and low values indicate non-vegetated land cover.
You can then 'slice' or categorise this NDVI raster into different bins or NDVI value ranges to create a classified map. If you have ground truth data (which I highly recommend collecting), this slicing can be more informed, allowing you to differentiate within classes which are vegetated (e.g. grasses and trees) and non-vegetated (bare ground and built-up areas).
In regard to improving classification results, keep in mind, that to calculate NDVI you need information for red wavelengths and near infra-red wavelengths in your image. The logic being that vegetation absorbs red and scatters infra-red, while things which are non-vegetated can scatter both. NDVI essentially is nothing more than a reorganisation of this red and infra-red information into a band ratio. So, calculating NDVI is not actually introducing new information into your classification. Rather, it is describing a relationship of red to infra-red for each pixel.
In my experience it has been useful to incorporate an NDVI layer during classification. But also do experiment with including other band ratios, such as Green/Red and Green/Blue.
Varun, you mention NDVI values ranging from -1 to +1; theoretically, I assume; I would be interested to hear if you, or anybody else, found NDVI values (DN) lower than say -0.2 in practice?
Hi Hein,
Yes, I do mean theoretically NDVI values should be bounded between -1 and +1. Also, getting values close to -0.2 or lower is rare, especially from satellite imagery that I have worked with covering quite varied landscapes. However, the lowest NDVI value I have seen in my data (from high resolution multi-spectral aerial photographs) is -0.61. It fell within a class I had designated for 'construction - metal/concrete'. But this is quite a rare occurrence and definitely an outlier even within the class. Minimum NDVI values generally hovered around -0.4 for the class.
It could be used for classifying vegetated and non-vegetated areas. The use of multi-temporal dataset (e.g. Landsat) can improve the results.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing