Parrot does not recommend the use of the protective lens cover for data acquisition during flight.
We work in the Arctic tundra and can't avoid the occasional dodgy landing due to the rough nature of the environment. There is a high risk of scratches on the lenses.
I was interested in how bad the transmissivity of the protective lens cover actually is, so I took some measurements in the lab:
Measurements were taken with an ASD 3 Field Spec and 8° field-of-view fore optic. The protective lens cover (filter) was mounted on a rotary platform between fore optic and an integrating sphere as an uniform light source. 5 angles were measured for each, horizontal and vertical field-of-view.
I repeated this for two filters shipped with our two Sequoia's (first batch?, spring 2016 - serial no. ending in 37 and 39). Results are highly consistent between both filters and both field-of-views. The drop in the extremes angles is slightly lower in the vertical field-of-view, due to smaller angles (data not shown).
The results suggest that the general impact of the filter on the light transmitted is relatively low (maximum reduction is -10%). However, the transmissivity of the filter is not uniform across wavelengths and drops towards the extremes of the field-of-view (up to -2.3% for the NIR band). The drop across the higher wavelengths is steady and the NIR band is particularly affected. The decline across the band is approximately -1.5% (lower to upper boundary). The other bands are fairly flat (slope +- < 0.05), the REG band gets away as it is only 10 nm wide.
My thoughts are:
- The drop in overall transmissivity and difference between bands should be no problem as this should be accounted for by radiometric calibration with a reference target.
- The drop across the NIR is a bit problematic, but should not matter when directly comparing flights that were flown with the filter applied.
- The reduced transmissivity towards the extremes of the filed-of-view is a reason for concern, but if the overlap is sufficiently high Pix4D and alike should mainly use the pixels from the centre of the image and the error introduced by the filter should be fairly small.
So, I'd be tempted to use the filter for the extra protection they provide ($3500 sensors) and live with the small error they introduce.
What are your thoughts?
The University of Edinburgh, Tundra Ecology Lab