Resolution is lacking, suggestions?

Joined
Nov 2, 2016
Messages
8
Reaction score
0
Age
57
I have a P4 and I'm trying to create an orthomosaic for a surveyor so he can do an overlay, align and then draw in details like curbs, sidewalks, manhole covers, etc.

I have been trying to utilize DroneDeploy Pro. I am doing large overlaps (80% on each) and have tried a variety of altitudes, but the end product just doesn't have the resolution he needs to be in the tolerances he wants. Edges are jagged and fuzzy when you zoom in. The actual photos are much clearer.

I export at .8in/px both as single image and as panels as GeoTIFFs.

I've also tried Skycatch with the same results.

Any suggestions?
 
I'm using the p3 pro I'm finding the same thing using the recommended arrangement of overlapping pics and straight down perspective with a height between 100 to 200 ft. Depending on which post you read the p3 camera is the same or the p4 is a bit sharper.

Here is a map I did and processed through drone deploy: http://drdp.ly/eCyj3I Does this map provide the detail you are looking for? I created the map by not only using the straight down pics but also taking many oblique pics. In all I used about 175 pics for 1 acre. Some parts of the pics are quite sharp while others parts is still blurry. Some of the pics I took quite low - 20 ft or so.
 
That is the detail I need.

So, to make sure I understand what appears to be a simple concept; I can add in photos from different heights and angles in the DroneDeploy processing engine and get a better final overlay. It doesn't have to be from a rather stagnant flight plan.
 
Another question, are you generating a GeoTIFF from the high altitude and then starting a new map and adding more detailed lower altitude photos for another GroTIFF and then adding the obligues or are you loading all altitudes and obliges at once and processing?

I took my 200 ft GeoTIFF and then started a new map with it as my base and added lower 40 foot shots, and it didn't appear to help.

Thank you for the help.
 
It is my impression that nadir photos from a height of 100 - 400 feet are most useful in agriculture scenarios where having a photo realistic orthomosaic is not desired. It is more useful to capture the general color of a row of plants and not an individual plant. Or if making an orthomosaic of a rock quarry there is no need to identify each rock. These types of orthomosaics can be captured with relatively few photos from a height between 200 - 400 ft. The object of the number of photos is to ensure that all areas are photographed at least 4 times - there is a graph in maps-made-easy which shows how many pictures cover a particular area.

To complete maps that are fairly photo realistic - that are sharp enough to make quite detailed measurements takes a different technique. At this point I do not know much about this. However from what I can tell take nadir pictures from much lower altitudes such as 20 - 50 feet helps - I'm not sure how low improves the results. In addition take oblique photos between 30 - 45 degrees seems to be the recommended angle. The one suggestion in taking oblique photos is to ensure the horizon is not included in the photo. Evidently the process cannot correctly account for the horizon.

I created the orthormosaic above by submitting all the photos - high altitude, low altitude, and obliques at the same time. I did notice that most areas on the map overlapped 10 or more times. I'm not sure if there is a limit to the number of overlapped photos or if the resultant image degrades if too many overlapped photos are submitted.

It would be nice to take some photos and the obtain an intial GeoTiff, then take some additional photos and create a more detail geoTiff. I do not know if the process can do this.
 
I have a P4 and I'm trying to create an orthomosaic for a surveyor so he can do an overlay, align and then draw in details like curbs, sidewalks, manhole covers, etc.

I have been trying to utilize DroneDeploy Pro. I am doing large overlaps (80% on each) and have tried a variety of altitudes, but the end product just doesn't have the resolution he needs to be in the tolerances he wants. Edges are jagged and fuzzy when you zoom in. The actual photos are much clearer.

I export at .8in/px both as single image and as panels as GeoTIFFs.

I've also tried Skycatch with the same results.

Any suggestions?

Resolution is not the problem. Expectations are the problem. The resolution is adequate for aerial surveying.
 
Resolution is not the problem. Expectations are the problem. The resolution is adequate for aerial surveying.

Well, when presented with images I gave to two surveyors, they said they were not adequate; therefore, the results I have been able to obtain up until a couple days ago were not good enough. After using the suggestions above and trying different techniques, I'm almost at a level they can use, but I'm still ironing out some wrinkles.

Thanks for the constructive suggestions. :p
 
It would be great if you do not mind sharing the results that are acceptable for these use cases and how you went about creating them.
 
I basically followed your suggestion. I did a relatively high pass at 250 feet, another at 150 and another at 100. I then flew manually at a much lower altitude over the ground reference points and any important features like driveways, sidewalks, structures. I also took obliques as you suggested. It takes several low altitude photos of each item, one of each important feature won't help. It seems to need three or four overlaps at that low altitude to draw in properly.

One issue I'm having right now is that things like power poles will be missing if the lower shots get too close to them. Trees will also be quite squirrely.

Even with this method, parts of roof lines are still not very crisp. I spoke with Drone Deploy, and they said they are still trying to hammer out the issues effecting roof lines.

I'm having to upload a crap load of photos in order to achieve what is needed. The problem is that a missing power pole won't go over well with a surveyor.

It's going to take a lot of refining the technique and practice to get a product surveyors are going to accept. Just flying at a couple of different altitudes isn't going to cut it.
 
That is the detail I need.

So, to make sure I understand what appears to be a simple concept; I can add in photos from different heights and angles in the DroneDeploy processing engine and get a better final overlay. It doesn't have to be from a rather stagnant flight plan.

You don't want photos from different elevations. You can do oblique images of buildings to generate a better 3D model. Just make sure the horizon is not visible in those photos. With dronedeploy I tend to over shoot the intended area so that I get proper overlap up to the property line.
 
For some reason, probably a very good one, long skinny poles - trees, telephone polls, are not included in the maps and models. As a funny example, I took pictures of my son trying to obtain a 3d model of him. My son, who is 5'6" was removed from the image and all that remained was his shoes.
 
Robert, I tried doing just oblique images and process it as structure. The roof lines come out better and things like poles aren't missing; however, when you overlay the 2D image over the top of a survey that has everything mapped, we found that some things like a sidewalk edge will be off by up to 6 inches.

Maybe I didn't take enough photos for it to get the correct accuracy.

How many oblique images do you think you take for a 1 acre lot with structures.

Due to trees, power lines, etc, some properties would be extremely difficult to line up those shots I'd have to imagine and nearly impossible for automated flight.
 
With the obliques you want to orbit the building. Enough to where you can get the whole building in the frame. I use the litchi app.

Just to be specific - when you orbit the structure - it is necessary to get the entire building such as a house in each pic. Or can each pic only have a portion of the house. It is just when you add up all the pics that the entire house is photographed somewhere during the orbit.
 
Just to be specific - when you orbit the structure - it is necessary to get the entire building such as a house in each pic. Or can each pic only have a portion of the house. It is just when you add up all the pics that the entire house is photographed somewhere during the orbit.

The more control points you get in the photograph the better the stitching. You want to get the whole building in each shot. The smaller the building the more detail.
 
ecolumbia999, yeah, that is what I encounter with utility poles and some trees.
The more control points you get in the photograph the better the stitching. You want to get the whole building in each shot. The smaller the building the more detail.

What are you doing to get all the ground information - sidewalks, driveways, manholes, water meters, culverts, etc?
 
I have tried both drone deploy and mapsmadeeasy and found the finished photos to be fuzzy and not acceptable. I use Pix4D and I can read the lettering on the manhole covers from 120'.

We just had a demo from TOPODOT. This package looks amazing for filtering vegetation and doing planimetric drawing of the infrastructure. Problem is that it uses microstation and we are civil3d. We are starting to have a lot of money tied up in software. The drone was the cheap part of the process! We did find an aerial photogrametry firm that will process our data for us to extract vegetation, create surface and generate planimetrics.
 
ecolumbia999, yeah, that is what I encounter with utility poles and some trees.


What are you doing to get all the ground information - sidewalks, driveways, manholes, water meters, culverts, etc?

To be honest I think your best bet is going to be manually stitching the images. All the software packages I've seen DD, MME and PIX4D all struggle to get the perfect stitch. I've uploaded the same photos to MME and DD and get different results with the same images.
 
I have tried both drone deploy and mapsmadeeasy and found the finished photos to be fuzzy and not acceptable. I use Pix4D and I can read the lettering on the manhole covers from 120'.

We just had a demo from TOPODOT. This package looks amazing for filtering vegetation and doing planimetric drawing of the infrastructure. Problem is that it uses microstation and we are civil3d. We are starting to have a lot of money tied up in software. The drone was the cheap part of the process! We did find an aerial photogrametry firm that will process our data for us to extract vegetation, create surface and generate planimetrics.

Are you flying at a single altitude?
 

Members online

Forum statistics

Threads
143,066
Messages
1,467,358
Members
104,935
Latest member
Pauos31