The NDVI measurement was derived in the 1970s as a way to take advantage of the NIR sensors on the, then new, Landsat satellites. It was developed to convert the amount of NIR light received at the satellite to the density of plants in each pixel of that satellite’s sensor array. Since each pixel corresponded to over 4 square miles on the ground this was useful to measure the scale of ecosystems. And it worked well.
Because the lighting conditions are constant from a satellite, and the cameras on the different satellites are highly calibrated, you can compare between images. A measurement of .5 on one image will mean the same amount of plant coverage as is found on an image with .5 on another NDVI.