The repo contains code for automated mapping of trees/plantation and farmland from 5m(spatial resolution) multispectral multi-temporal (LISS IV) data. This code has been written in R and the end goal is creation of a shiny webapp for viewing and mapping vegetation areas in the TIF files uploaded. Some sample images can be downloaded from this link
- Images are acquired from Indian Remote Sensing Satellite Resoursesat -1,2 with LISS-IV sensor.
- It provides multi-spectral data with three bands with spatial resolution of 5m at nadir. The given imageries are band stacked FCC (False Colour Composite) as (Band 1- Near Infrared, Band 2- Red and Band 3- Green).
- It has 10 bits radiometric resolution with the swath of 70 km and the revisit time of 5 days.
Vegetation reflects more NIR than other wavelengths but water absorbs NIR, therefore the lighter areas with high reflectance values are likely to be vegetation and the dark blue, low reflectance value areas, likely to be water. Chlorophyll is the key factor in reflectivity of vegetation. It absorbs strongly in red, giving rise to our visible observation that healthy vegetation is green. But chlorophyll reflects most strongly in very near-infrared (VNIR) just beyond the visible.
All the following has been generated by using the raster package in R, the code for which can be found in code.R
The red and green wavelengths are represented in the above image. We can get a rough approximate of the vegetative areas in this image
Over here, the NIR band is represented with blue colour and we can see the vegetation that has reflected the near infrared light.
The Normalised Difference Vegetation Index (NDVI) is a widely used vegetation index that quantifies vegetation presence, health or structure. It is calculated using the Near Infrared (NIR) and Red bandwith of the spectrum. Healthy vegetation reflects light strongly in the NIR part of the spectrum and absorbs light in red part of the visible spectrum for photosynthesis. A high ratio between light refected in the NIR part of the spectrum and light reflected in the red part of the spectrum would represent areas that potentially have healthy vegetation. It is worth noting that different plant species absorb light in the red part of the spectrum at different rates. The same plant will also absorb light in the red band differently depending on whether it is stressed or healthy, or the time of year. It is often used over large areas as an indication of land cover change. The NDVI ratio is calculated using (NIR - Red) / (NIR + Red). For example, a pixel with an NDVI of less than 0.2 is not likely to be dominated by vegetation, and an NDVI of 0.6 and above is likely to be dense vegetation.
By taking 0.2 as a threshold NDVI value to categorise the pixel as vegetation, we can remove all the pixels with NDVI < 0.2
Plotting a histogram of NDVI values of the image, giving an approximation of vegetation in the given image
We see that the graph is right skewed, meaning the image represents healthy vegetation
