Many of our customers are interested in tracking changes in plant health, so naturally the question we get asked often is which of our multispectral filters work the "best". We also get asked why the index values are different when using similar filters, such as our RGN and OCN. Read on to learn more:
Looking at the below graph you can see the irradiance of both healthy and dead leaves. The ambient light was incandescent bulbs, as they produce a light spectrum similar to sunlight.
When we are looking to find the differences between two objects (healthy vs dead) we are looking for where in the above reflective signatures (graph lines) there is the biggest separation (gap). The larger the separation the more contrast exists when looking at that particular spectral range.
For plants, the largest difference between healthy and dead leaves is typically red light (~640-700nm) and near infrared light (~800-900nm). As you can see in the above graph, these regions are where the green and red lines differ the most.
Below is the same graph but with the RGN and OCN filter transmission bands (FWHM) included:
Notice how the near infrared (NIR) channels are similar in contrast, but the red channel is much larger than the orange. This is why NDVI and most vegetative indices use red instead of orange compared against the NIR. Both can be used, but the reflected red light will have more contrast with the near infrared light.
Something also interesting is that there is very little difference in reflected green light between healthy and dead leaves. Dead leaves just reflect more red light, and when green and red are mixed you get brown (the common color for dead vegetation).
Now let's take a look at some images from our Survey3 RGN and OCN cameras (click to view):
We arranged some healthy and dead leaves on a white cotton sheet. The ambient light was the incandescents shown in the graphs above.
The leaves were grouped together on the far right and measured with a spectrometer to obtain the spectral signatures shown in the above graphs.
When using camera images for remote sensing the pixel values can be processed with an index formula to tell us information about the scene. For customers looking at plants the multispectral formula NDVI, Normalized Difference Vegetation Index is commonly used to explain a range of vegetation health. NDVI ranges from -1 to +1, but plants tend to range from 0.2 to 1. A lower NDVI value represents dead or less healthy and a higher NDVI represents medium to good health.
The NDVI formula is a normalized ratio between the reflected near infrared (NIR) and a visible light color (VIS). As we have seen above, red works the best due to it reflecting less light than orange for healthy vegetation.
Here are the NDVI formulas using red or orange light:
Notice that the NDVI legend has the same minimum and maximum values for the RGN and OCN images, so the green-yellow-red color gradient (lut) is the same.
We then selected the regions of the spectrometer scan data that pertained to the RGN and OCN channels (FWHM) and computed the NDVI values. Those computed NDVI values are listed in the black boxes over the grouped leaves to the frame right, and below:
CAMERA FILTER
DEAD LEAF NDVI
HEALTHY LEAF NDVI
RGN (Red, Green, NIR)
0.479135
0.791951
OCN (Orange, Cyan, NIR)
0.618256
0.790290
The green healthy leaves were the same computed NDVI value regardless of the camera filter used, which is confirmed by the green color in the NDVI Survey3 camera images.
Also as expected, the NDVI value for the dead leaves is different depending on which spectral region is used:
When you use the red channel from the RGN filter model camera, the NDVI value for the dead leaves is lower. This is supported by the RGN image showing a lower NDVI value, and thus an entirely red color for the dead leaves. For these particular leaves the RGN camera will show more contrast when using the NDVI index.
When you use the orange channel from the OCN filter model camera, the NDVI value for the dead leaves is higher. This is supported by the OCN image showing a higher NDVI value, and thus a mixture of yellow and orange pixels. For these particular leaves the OCN camera will show less contrast when using the NDVI index.
After reading the above you should now understand why the RGN filter has traditionally been the "best" option for vegetative indices like NDVI. It tends to produce more contrast, and thus it is easier to see changes in plant health sooner.
You can also see that index values produced by multispectral index formulas such as NDVI are fully dependent on where and how much of the light spectrum you sample. It is rare for two instruments to produce the same index values unless they are sampling the exact same region of the spectrum.
CONTACT UStoday to talk to a multispectral remote sensing expert. We are happy to assist with explaining our technology and helping you select the best solution for your project.
