Welcome, Commercial Drone Pilots!
Join our growing community today!
Sign up

Issues processing similar ground

b-c

Member
Joined
Jan 18, 2018
Messages
5
Reaction score
2
Age
40
Location
Central California
Website
www.sclr.org
Recently we went to fly / survey / map some ag ground primairly for measurements vs specifically AG related activity...

Point being the ground was freshly plowed before being planted, software was unable to generate a point cloud for modeling (Survey work mostly)
What do others do to get the best possible outcome on large properties with very similar features in each image?

we had some ground points setup and flew at different altitudes 300' with an inspire 2 and X5S
small acreage ~100 (specific area of interest was probably less than 10)

with all that said - pretty much no processing packages - pix4d / precission mapper or others were able to differentiate the images due to the images being too similar...

----
Side bar - also found the inspire tagging images with both barometer (correct altitude agl) and gps - showing ~ 200' higher than actual flight which if that # was used for an investigation, vs flight logs - would get us in trouble for being over 400' limit - we were not but the # was bad.

Pix4d didn't grab barameter alt it grabbed gps at and ~40% of images were tagged with incorrect alt... just fun...
 

ArrUnTuS

Well-Known Member
Joined
Jan 5, 2018
Messages
360
Reaction score
144
Recently we went to fly / survey / map some ag ground primairly for measurements vs specifically AG related activity...

Point being the ground was freshly plowed before being planted, software was unable to generate a point cloud for modeling (Survey work mostly)
What do others do to get the best possible outcome on large properties with very similar features in each image?

we had some ground points setup and flew at different altitudes 300' with an inspire 2 and X5S
small acreage ~100 (specific area of interest was probably less than 10)

with all that said - pretty much no processing packages - pix4d / precission mapper or others were able to differentiate the images due to the images being too similar...

----
Side bar - also found the inspire tagging images with both barometer (correct altitude agl) and gps - showing ~ 200' higher than actual flight which if that # was used for an investigation, vs flight logs - would get us in trouble for being over 400' limit - we were not but the # was bad.

Pix4d didn't grab barameter alt it grabbed gps at and ~40% of images were tagged with incorrect alt... just fun...

What level of overlap both front and side?

In the scenarios that you have described, such as large forest masses where vegetation is predominant, in this case it would be the soil, Pix4D recommends increasing the overlap, decreasing the height to capture more detail, the time of day is important to avoid reflections, or processing it with better quality cameras so that by providing increased quality and detail in the images the program is able to make the joining process.
 
  • Like
Reactions: b-c

b-c

Member
Joined
Jan 18, 2018
Messages
5
Reaction score
2
Age
40
Location
Central California
Website
www.sclr.org
80% on both... also ran tests at 200' - think it was pix4d capture running both attempts...
default settings - might have increased on second run. - also used the newer polygon setting.
(seems less overlap would have been better)

seriously with the X5S the detail is extremely high

didn't have many issues using an Iris+ with Canon S110 (I think)

typically flew the iris+ at 100'-180' just didn't have good coverage...
and battery estimate was at best 10m vs 20-25 with inspire and bigger camera.

we did consider flying at 100' but another 2 sets of batteries is possible just odd...
 

ArrUnTuS

Well-Known Member
Joined
Jan 5, 2018
Messages
360
Reaction score
144
The truth is that it has never happened to me but I remember the article that published Pix4D and recommended the 85% overlap in both, although it referred to flights over the jungle that in the end I understand that it is the same because in the end the problem is that it does not find overlapping points.

Doing a quick search now I've found this:

85% frontal and 70% side overlap for forests, dense vegetation and fields.

I may have changed or got confused but I've kept the idea of 85% in both.

Elsewhere I have read that the recommendation was to fly when sunlight is low. At dawn or dusk so that the projected shadows cast shadows even on the smallest objects and allow for a successful union of the images. Of course with enough light to get valid results.

Can you please upload a photo of the problematic terrain so that we can get a visual idea of the problem?
 

b-c

Member
Joined
Jan 18, 2018
Messages
5
Reaction score
2
Age
40
Location
Central California
Website
www.sclr.org
Sure... this one is reduced but you get the idea...
24pct-9.5mb.JPG

typically run with lower shadow either 11-2pm range and good light...
have also read most of those articles as well.
 

ArrUnTuS

Well-Known Member
Joined
Jan 5, 2018
Messages
360
Reaction score
144
Sure... this one is reduced but you get the idea...
View attachment 100

typically run with lower shadow either 11-2pm range and good light...
have also read most of those articles as well.

It's complicated stuff. Besides, there's not a stone, the farmer's dream and the drone pilot's nightmare :eek:

Having to go back to the field to make flights again is the worst thing that can happen to you. Apart from losing the images.......

I imagine that if you do not do more tests that can give a positive result, until what is planted there is not born, there will be no solution.......

Please comment on whether you accomplish anything.
 

Mike Nevins

Well-Known Member
Joined
Jan 19, 2018
Messages
130
Reaction score
37
Location
Grand Junction, Colorado
Website
www.mountainskyaerial.com
I guess I'm confused, why would you want to model flat ground in a point cloud? As far as whats the best time to fly for mapping, its when you have ideally an overcast day with no sun. No sun, no shadows. Evening or Sunrise would both be bad for mapping. Shadows are bad for mapping, surveying and for Crop Scouting. Just my experience.
 

ArrUnTuS

Well-Known Member
Joined
Jan 5, 2018
Messages
360
Reaction score
144
I guess I'm confused, why would you want to model flat ground in a point cloud? As far as whats the best time to fly for mapping, its when you have ideally an overcast day with no sun. No sun, no shadows. Evening or Sunrise would both be bad for mapping. Shadows are bad for mapping, surveying and for Crop Scouting. Just my experience.

If you are asked to make measurements you have to create a 3D model. If you want to obtain an orthophoto it is necessary that the program be able to make the union. This is where the problem arises, that the software is not able to make the union, first step to get the rest of the results.
 

Jesse G

Well-Known Member
Joined
Jan 3, 2018
Messages
59
Reaction score
37
Age
37
Recently we went to fly / survey / map some ag ground primairly for measurements vs specifically AG related activity...

Point being the ground was freshly plowed before being planted, software was unable to generate a point cloud for modeling (Survey work mostly)
What do others do to get the best possible outcome on large properties with very similar features in each image?

we had some ground points setup and flew at different altitudes 300' with an inspire 2 and X5S
small acreage ~100 (specific area of interest was probably less than 10)

with all that said - pretty much no processing packages - pix4d / precission mapper or others were able to differentiate the images due to the images being too similar...

----
Side bar - also found the inspire tagging images with both barometer (correct altitude agl) and gps - showing ~ 200' higher than actual flight which if that # was used for an investigation, vs flight logs - would get us in trouble for being over 400' limit - we were not but the # was bad.

Pix4d didn't grab barameter alt it grabbed gps at and ~40% of images were tagged with incorrect alt... just fun...
With your main interest beaing about 10 acres I would try staying at around 300' since that'll give you about .79" GSD with the 15mm lens but I would increase front overlap to 90% and side to 85% simply because the more data you have the better chance of stitching. Then I would consider setting out some targets in a grid every few hundred feet so 1 or 2 are visible in each image. I use 24" or 30" checkerboard targets printed out on a lightweight vinyl from my local copy center, I did testing with plain paper first and it worked as well and a 24" print is only 24 cents. I either tape them down to concrete/pavement or tape them to 1/4" press board and lay them out to shoot with my RTK system but have tried applying them to contoured ground simply shoveling dirt onto the edges to keep them in place with good results so maybe you could even lay them on the rows to follow the terrain. Even if your not using an RTK system the targets could give you the needed tie points to stitch. Will mean a decent amount of walking the rows of the field to place and pick up the targets but may be worth a shot. Once when mapping a large dessert area with lots of areas of bare featureless ground I added about a dozen targets on top of the 10 I already had and drove my truck throughout the areas to add texture to finally get the images to stitch...

The switching elevation tagging is something I haven't heard about on the I2 but the gps elevation tags have been bad on all DJI platforms. There is a work around for that posted on the inspire pilots forum here ...
Sorting out DJI's bad GPS altitude
But that doesn't addres the 200' difference in tagging from photo to photo...I've never seen that on my old P3P, I1 Pro or mavic
 
  • Like
Reactions: ArrUnTuS

ArrUnTuS

Well-Known Member
Joined
Jan 5, 2018
Messages
360
Reaction score
144
With your main interest beaing about 10 acres I would try staying at around 300' since that'll give you about .79" GSD with the 15mm lens but I would increase front overlap to 90% and side to 85% simply because the more data you have the better chance of stitching. Then I would consider setting out some targets in a grid every few hundred feet so 1 or 2 are visible in each image. I use 24" or 30" checkerboard targets printed out on a lightweight vinyl from my local copy center, I did testing with plain paper first and it worked as well and a 24" print is only 24 cents. I either tape them down to concrete/pavement or tape them to 1/4" press board and lay them out to shoot with my RTK system but have tried applying them to contoured ground simply shoveling dirt onto the edges to keep them in place with good results so maybe you could even lay them on the rows to follow the terrain. Even if your not using an RTK system the targets could give you the needed tie points to stitch. Will mean a decent amount of walking the rows of the field to place and pick up the targets but may be worth a shot. Once when mapping a large dessert area with lots of areas of bare featureless ground I added about a dozen targets on top of the 10 I already had and drove my truck throughout the areas to add texture to finally get the images to stitch...

The switching elevation tagging is something I haven't heard about on the I2 but the gps elevation tags have been bad on all DJI platforms. There is a work around for that posted on the inspire pilots forum here ...
Sorting out DJI's bad GPS altitude
But that doesn't addres the 200' difference in tagging from photo to photo...I've never seen that on my old P3P, I1 Pro or mavic

I hadn't thought of using targets, what a good idea. :) I've been lucky and never had those problems, but if I have them now I know how to solve them :D

Doubtless, aren't those targets too big? I, after doing the calculations of the recommended size and considering the 400 foot ceiling, gave me a size of 16 inches (40cm). The targets I have are 20 inches (50cm), more than enough.

Don't you always use the targets on a hard support? just stick them on the floor and only once in a while when you use hard support? I have the vinyls glued to galvanized steel sheets. The truth is that they are kilos to move around when you take them all at once but good.
 

Jesse G

Well-Known Member
Joined
Jan 3, 2018
Messages
59
Reaction score
37
Age
37
I hadn't thought of using targets, what a good idea. :) I've been lucky and never had those problems, but if I have them now I know how to solve them :D

Doubtless, aren't those targets too big? I, after doing the calculations of the recommended size and considering the 400 foot ceiling, gave me a size of 16 inches (40cm). The targets I have are 20 inches (50cm), more than enough.

Don't you always use the targets on a hard support? just stick them on the floor and only once in a while when you use hard support? I have the vinyls glued to galvanized steel sheets. The truth is that they are kilos to move around when you take them all at once but good.
You can definitely use smaller targets but I use the larger ones since photoscan offers automatic target detection and requires a decent amount of pixels for each target to be recognized and automatically placed accurately. When the targets are completely flat on a hard support it does reduce the likelihood that in some images you might not be able to see the center of the target and autodetection might fail but thats seemed to be the only issue. And for this use as it seems like the targets could really just be used for pixel variation it shoildnt be an issue. However shooting the targets in with an RTK system will help keep the elevation of the model accurate as with such similar features on the ground photogrammetry software does tend to struggle. I first noticed what difference variation makes when mapping a large subdivision in which half of the roads were brand new paving and half was older and had been cracksealed. The new roads came out horrible do to the lack of features on the fresh black surface and the cracksealed roads came out almost perfect in the areas with the most sealed cracks since they provided tie points along the cracks.
 
Last edited:

ArrUnTuS

Well-Known Member
Joined
Jan 5, 2018
Messages
360
Reaction score
144
You can definitely use smaller targets but I use the larger ones since photoscan offers automatic target detection and requires a decent amount of pixels for each target to be recognized and automatically placed accurately. When the targets are completely flat on a hard support it does reduce the likelihood that in some images you might not be able to see the center of the target and autodetection might fail but thats seemed to be the only issue. And for this use as it seems like the targets could really just be used for pixel variation it shoildnt be an issue. However shooting the targets in with an RTK system will help keep the elevation of the model accurate as with such similar features on the ground photogrammetry software does tend to struggle. I first noticed what difference variation makes when mapping a large subdivision in which half of the roads were brand new paving and half was older and had been cracksealed. The new roads came out horrible do to the lack of features on the fresh black surface and the cracksealed roads came out almost perfect in the areas with the most sealed cracks since they provided tie points along the cracks.

Yes, you have to be careful when putting them on and making them as horizontal as possible to avoid problems. Matte vinyl to avoid reflections and choose well, whenever possible, where to put them.

The other reason to opt for rigid supports is to prevent them from moving with the wind, otherwise a fluorescent spray might also work but then leave traces that although in most cases is not important in areas already finished is a problem.

What you're saying is another interesting option, I'm writing it down :)
 

New Threads

Members online

Forum statistics

Threads
3,082
Messages
29,262
Members
4,438
Latest member
adokora