Need some help with Dronedeploy workflow and NDVI interpretation

Joined
Jul 14, 2016
Messages
22
Reaction score
9
Location
Norway
I've got a Mapir Survey2 NDVI-camera on one of my Solos, and after some initial testing this summer I spent a few weeks exploring 3D-mapping and general photo flights. My cousin has a wheat farm though, and he's fairly interested in new technology. So I paid him a visit on Sunday to see what we could make out of a couple of his fields.

The trigger on the Survey2 is set to 0.5sec, so I usually get an image every to, sometimes three seconds.
I don't have any means of calibraton. It was a sunny day, with some clouds, and I don't know how they affect the NDVI.

In the first run which was at 100 meters AGL and 5ms speed, the shutter was set to 1/1000 as default.
For the second run I made the mistake (again) to adjust the ISO from default 100 to 200. I know this messes with the calibration, but I have no idea whether that can cause the differences I will shortly show you.
The 50m AGL images was taken some 3hours later in the day, but it was mainly a sunny day.
I did not shoot these in RAW, as my memory card was about to fill up.

This is the area in question. Taken with my QX1. The new wheat in this field is just 10-15 cm tall.
Here is a link to the Dronedeploy 3D-model if you wish to look at it.
http://drdp.ly/6vbPnS

skinnes01.PNG skinnes05.PNG

Now this is how the 50m AGL scan comes up. I've changed filter type to RGN and I use SAVI as that seems to be recommended for areas with much soil and little vegetation.
skinnes02-50m8ms.PNG

Now for the 100m AGL scan, same settings here, but note that the accuracy RMSE figure is so much worse. Orthomosaic resolution should be worse due to the elevation.
skinnes03-100m5ms.PNG

So... these two maps are so different that I'm curious to what I'm doing wrong. Is it the lack of calibration? Is it the ISO difference between the two runs? Here I have the two maps compared side by side - with the suggested scale settings of Dronedeploy.
skinnes04-compare3.PNG

Here I try to change the scales of both to match, first trying to match the 50m AGL map to the settings of the 100m AGL map.

skinnes04-compare1.PNG

And here I try to compare the settings of the 100m AGL map with the 50m one

skinnes04-compare2.PNG

and I'm like..... what the heck is going on...

ANY input is appreciated on this...Why am I seeing this huge difference? Are either of the maps useless? Both?
I'm probably going up there again next weekend, but I would really like to figure out what makes for such a difference in the maps and what I should do differently/better on the next run.
I know Dronedeploy isn't the solution with the best lut, but it's the least hassle solution and it must be working to some extent, otherwise nobody would have used it? :).

Thanks for your time :).
 
Joined
Jun 4, 2015
Messages
2,911
Reaction score
1,699
Age
64
I've got a Mapir Survey2 NDVI-camera on one of my Solos, and after some initial testing this summer I spent a few weeks exploring 3D-mapping and general photo flights. My cousin has a wheat farm though, and he's fairly interested in new technology. So I paid him a visit on Sunday to see what we could make out of a couple of his fields.

The trigger on the Survey2 is set to 0.5sec, so I usually get an image every to, sometimes three seconds.
I don't have any means of calibraton. It was a sunny day, with some clouds, and I don't know how they affect the NDVI.

In the first run which was at 100 meters AGL and 5ms speed, the shutter was set to 1/1000 as default.
For the second run I made the mistake (again) to adjust the ISO from default 100 to 200. I know this messes with the calibration, but I have no idea whether that can cause the differences I will shortly show you.
The 50m AGL images was taken some 3hours later in the day, but it was mainly a sunny day.
I did not shoot these in RAW, as my memory card was about to fill up.

This is the area in question. Taken with my QX1. The new wheat in this field is just 10-15 cm tall.
Here is a link to the Dronedeploy 3D-model if you wish to look at it.
http://drdp.ly/6vbPnS

View attachment 4232 View attachment 4238

Now this is how the 50m AGL scan comes up. I've changed filter type to RGN and I use SAVI as that seems to be recommended for areas with much soil and little vegetation.
View attachment 4233

Now for the 100m AGL scan, same settings here, but note that the accuracy RMSE figure is so much worse. Orthomosaic resolution should be worse due to the elevation.
View attachment 4234

So... these two maps are so different that I'm curious to what I'm doing wrong. Is it the lack of calibration? Is it the ISO difference between the two runs? Here I have the two maps compared side by side - with the suggested scale settings of Dronedeploy.
View attachment 4237

Here I try to change the scales of both to match, first trying to match the 50m AGL map to the settings of the 100m AGL map.

View attachment 4235

And here I try to compare the settings of the 100m AGL map with the 50m one

View attachment 4236

and I'm like..... what the heck is going on...

ANY input is appreciated on this...Why am I seeing this huge difference? Are either of the maps useless? Both?
I'm probably going up there again next weekend, but I would really like to figure out what makes for such a difference in the maps and what I should do differently/better on the next run.
I know Dronedeploy isn't the solution with the best lut, but it's the least hassle solution and it must be working to some extent, otherwise nobody would have used it? :).

Thanks for your time :).
I would think that the first order of business is to 'ground truth' against the images. Walk areas of the field with the land owner and compare visually the areas of good and marginal plant areas. Compare that against the above images and see what best matches what you find on the ground. This would be very useful and I would be interested in your results as well.
 
Joined
Jun 18, 2015
Messages
3,113
Reaction score
1,646
Location
Chouston, Tejas
I thought the sun's angle would effect how these images reveal properly. Seems your shadows are long (trees and buildings) and could effect how the shorter plants are "over shadowed" by taller plantings. I was surprised that the shadows seemed stationary from flight to flight, since you mentioned 3 hours span between flights.

Just curious, what time of day and at what latitude?
 
Joined
Jul 14, 2016
Messages
22
Reaction score
9
Location
Norway
First flight around 13.30, last flight a little after 16.00. Lon/lat 60.1N/9.7E, just north of Oslo. The shadows are indeed long, but I think the main issue is that the plants are so tiny at this stage. The field is very open and buildings and trees casting a shadown are only on the right side of the mapped area.
 
Joined
Jun 18, 2015
Messages
3,113
Reaction score
1,646
Location
Chouston, Tejas
Sorry, now I understand. The image outside the ortho area is a static image that was overlayed....my apology. Very misleading.....
 

New Threads

Members online

No members online now.

Forum statistics

Threads
12,862
Messages
146,422
Members
15,345
Latest member
valkata752