Problem with webodm generated DSM/DTM

Hello, I’m having problems with DSMs/DTMs generated with webodm, it shows elevations where there are not… compared to the same data sets processed with pix4d and drn2map and I’m not being able to succesfully post process the DSMs/DTMs in Qgis… any idea why??? Data sets where captured with a Mavic pro

Hi @Peke, could you post a few screenshots comparing the DSMs between ODM and Pix4D / Drone2Map?

Sure! @pierotofy Here you go a screenshot from both dsm opened in qgis, please see the differences in the layer attributes section. Thanks for your help! :slight_smile:
I’m using the latest webodm version and dataset was captured with a mavic pro

So they show different elevation values?

Do you have the flight altitude and ground altitude values so that we can assess whether ODM or Drone2Map is off? Or perhaps copy/paste the EXIF values of a few of your JPEGs so we can understand your dataset a bit better.

would it have to do wih photo distortion?? am I supposed to calibrate the photos of the mavic pro before feeding to webodm?? Thanks for your help :slight_smile: Here’s a screenshot from a exif data viewer of all the photos :slight_smile: image

Fligth heigth was 40 meters using pix4d fligth app :slight_smile:

If the aircraft altitude was 1954m (as recorded in the EXIF tags), and the flight height was 40m above ground, then I would expect the elevation values to be around 1914m (1954m - 40) (which is what OpenDroneMap computed?). Looks to me that Drone2Map might be off with its values in the 1870 range?

But… can you see that webodm shows me a depresion where drn2map shows elevation right in the center of the dsm?? i ran exactly the same photos on both programs and drn2map gives me the accurate terrain representacion as seen in person. Do I have to calibrate my photos before feeding to webodm?

Here is another process… I`d appreciate your help on this issue. :slight_smile: And some information on the atributtes from the same spot on both dsm on the left panel from qgis
Screenshot_20181011-152348_Pix4Dcapture WEBODM%20PROBLEM%20DSMS

Ah, I see what you mean.

Could you upload your point cloud output somewhere like Google Drive or Dropbox? Perhaps there’s a problem there.!AnHDUrz49X3kghOdC0mh9Q4KMfka
Here is the laz file!AnHDUrz49X3kghT7R0yL3bDAxkYh
@pierotofy my main concern here is if i`m doing something wrong my self, please answer if i need to calibrate the mavic pro photos before feeding webodm. I need to know if the problem is on my self because i intend to work only on webodm in the future and contribute to the project as soon as i make my first profit with it but i need to be sure that i can rely on webodm

It looks like there might be a “bowl shape” effect on your point cloud so perhaps you should try to undistort your images. I’ve never done much work with the mavic, so I don’t know the amount/type of distortion of the camera. ODM tries to undistort the images using a photogrammetric process, but not all distortion models can be accounted for that way.

Classic dôme effect from gimbal and near perfect flat flight. Put your camera at 45 degrees you will not have this problem.
In other hand, most dome effect are related to camera calibration

Correct. Fly your camera at an angle off nadir. 10-15 degrees off nadir is fine and will give better orthos, or you can fly a nadir flight and then orbit with camera at 45 and combine the two sets of images.

I’m curious as to how Drone2Map is able to process the dataset… do they just undistort the images prior to processing here?

The “fly the dataset again” might be a good workaround, but not ideal.

Agreed. We need to do a deep dive into how we’re doing automated image calibration so that we don’t require this kind of flying to get good DTM results.

Thanks everyone for your help on the mather. @smathermather @pierotofy @kikislater i actually have datasets already processed in webodm previously to this ones, i have dsms from a 3dgrid fligth with pix4d app where photos are captured at 70º and also i have another data set flown in a circular pattern from the same place, and also another one where i combined both 3d grid fligth and circular fligth trying to get a better 3d reconstruction and i tried all those DSMs outputs in qgis and i'm having the same problems processing in qgis... This is really getting to my nervs haha, i really aprecciate your time and effort on this project, but please help me out on this one, how may i calibrate my images properly, what program should i use that you know does it rigth?? ive spent ours reading about webodm but i`m running out of time to figure out how to get a proper product, i tougth since mavic pro is so popular i just had to feed the images just like pix4d, drn2map and all those other comerccial products, i soon want to launch dsm based products for customers but i cannot still rely on webodm to do the work… please help me on this one…!! :slight_smile:

I haven’t had these effects persist if I combine angles, @Peke, though I have no doubt this is what you are seeing.

Unlike closed source projects, we can take contributions to get things up to product level, so if you have time or funds to contribute to help us with calibration, we are happy to help. Otherwise, this will need to wait until we have time or funds to address.

One thing you can contribute now, irrespective of time and funding is an issue at documenting with source images and comparable datasets processed separately that demonstrate something closer to the correct answer. And thanks for bringing this to our attention!

I currently cant support the project with funds.. i was hoping to start attending clients and be able to offer solutions trhough webodm, however id love to contribute once i get comercial grade results if you point me out to a tutorial on how to calibrate my photos before feeding them to webodm, i`m not very skilled in informatics but i like to learn, and thanks a lot for the support and creating odm. And i did opened a thread at github about this but @pierotofy closed it… :frowning:

i`ll try to open a new one though.