Effects of skewed imagery when processing?

[doktorinjh]

I am trying to process some data from a DJI Inspire and the images were taken at about a 20 degree angle to the direction of travel, as in below:

The left picture is what you think the photo alignment should look like, but the images we have were taken like the diagram on the right. Does the photo processing software take this into account when it tries to align and match the photo pairs? The exif data from the Inspire has some other potential errors, such as the Omega fluctuating from 90 to 0 to -90 and back, throughout the images, causing some errant results. Photoscan has had decent results, but Pix4D is having a fit trying to process the data.

The images were received from a subcontractor and I'd also like to help them correct the skewing, so please let me know if you think this could be due to a compass misalignment or other issue. It has happened to me before using a P2, but I think I just lived with the results at that time. Thanks!

 

[Meta4]

Photoscan has had decent results, but Pix4D is having a fit trying to process the data.

I don't know about Pix4D but Photoscan will handle any orientation of the images - it's just looking for points to match.
This sounds like a question for Pix4D.
I usually get skewed images at the ends of rows where the Phantom is turning and Photoscan has no problem with that.

let me know if you think this could be due to a compass misalignment or other issue.

I can't see how compass issues would do that.
I'd suspect that the Inspire gimbal/camera was rotated to that orientation.

 

[doktorinjh]

I don't know about Pix4D but Photoscan will handle any orientation of the images

Photoscan definitely has been producing better models and orthos for us recently and we've been doing side-by-side comparisons with lots of projects. My thought was that perhaps the image comparison was done from the "outside-in" and that somehow the skewed alignment was causing issues, but then you would have worse alignments during the turns and perhaps none using a double-grid pattern, so I don't know.

As for the crabbed flights, I've seen it before and I know that it's not a gimbal or camera issue. I would get it when I had a Canon S100 mounted to a non-gimbal frame using a hard aluminum mount to the chassis. My P2 would show a crabbed movement like the right side of the picture and then a more normal alignment coming in the opposite direction. I'll try to get an image that shows what that looked like. We saw it in the field and also in the post-processing.

 

[kitefiter]

I just flew a mission where the first 30 photos were aligned with the flight. The next 75 were skewed 45 degrees! I was flying with 85% forward overlap and 80% side overlap. Pix4d had no problem processing the photo. The pixel matching numbers per photo were just fine and the finish ortho looks great.

Still curious about the crabbing behavior though.

 

[doktorinjh]

Still curious about the crabbing behavior though.

Out of curiosity, what are you flying? Have you ever seen that behavior before?

 

[kitefiter]

P3A. And no, I have not seen this behavior before. I have seen skewed images on the turn only sometimes.

 

[doktorinjh]

I have not seen this behavior before

Here are 4 different projects where there was skew. These are old projects using a P2, but I think the issue is the same. In the first two pics, the skew is consistent in the northbound legs, but less pronounced in the southbound. The other 2 appear to be off axis in both directions. There was likely little to no wind when these were flown because I was very wind-shy when these were flown. Thought I would share.

 

[RocketBrew]

I see this frequently with both P3P and P4P due to the wind. DroneDeploy handles processing them without any problems.

I haven't seen the random skewing issues.