Messages

61

General / New PC system and scaling of performance

« on: May 04, 2021, 03:42:12 PM »

Hello.

It's about a new computer system for our work.
Which computer configuration is "good" is asked every few months.
I now simply trust the site https://www.pugetsystems.com, among others.


Two things.

1.)
We have an offer for a very expensive (5 digit) workstation (with AMD TR 3970X processor, 256GB Ram and Quatro RTX6000 graphics card).
However, some of our projects can easily be divided into chunks.
If we would buy 3 single "super" computers for the same price of this workstation, we would "really" be 3 times faster, right?
(11900k, 128GB Ram, RTX3070/3080)

2.)
Does Metashape really scale that badly to larger computer systems?
Putget suggests so.

An 8CORE Intel 11900k is faster than a Threadripper system?
An RTX3070 is only 5% slower than an RTX3090?

You'd think Metashape would be good at dividing into tasks and distributing them across CPU cores.

62

Feature Requests / Question about the "job organiser"

« on: May 02, 2021, 12:24:33 PM »

Hello.

I still have problems understanding the "Job Manager".
The arrow keys to move the priority have no effect on me.
Regardless of whether the current job is paused or not, I cannot move the "underlying" job forward.

Example:
A calculation that I start and then put in the background ends up under "Jobs".
Now I can open another project and work in it.
All changes (even just assigning new photos) now also end up in the job manager as a subordinate job.

Unfortunately, I can't change the order there with the arrows.
Even if I pause the first job, I can't get the subordinate job up or started.

Is this the right way?
It would be more flexible if I could briefly interrupt the "long two-day" job to do short calculations on the next one.

63

Bug Reports / Re: Errors when merging point clouds with multiple classifications.

« on: April 27, 2021, 11:09:33 PM »

Incredible, that was fast.
...and works!

64

Bug Reports / Errors when merging point clouds with multiple classifications.

« on: April 27, 2021, 09:55:15 PM »

Ok, I think I have mentioned this before.

When merging point clouds with multiple classifications, those are deleted.
(Or also merged).

The merged point cloud only has the classification: "Never classified"

Can this please be changed!?

65

Bug Reports / Re: Not a bug, but a misdirection?! Error estimation of the UAV images...

« on: April 24, 2021, 07:00:37 PM »

Until now, this has never been a problem.
Here it occurred for the first time.
However, I would have to "check" 1200 pictures soon.
It is not only the "good" or only the "bad pictures". It is the difference between the both values...

I'll wait for the next projects..

66

Bug Reports / Re: Not a bug, but a misdirection?! Error estimation of the UAV images...

« on: April 24, 2021, 04:55:41 PM »

OK, thank you!

It's probably not so much a time delay, since most of the values are correct.
In this case, the RTK connection often failed. RTK ON/OFF/ON etc...
I think some pictures were not determined exactly...

67

Bug Reports / Not a bug, but a misdirection?! Error estimation of the UAV images...

« on: April 19, 2021, 10:00:55 PM »

Hello.

During a normal drone flight, it often happens that the GNSS correction fails for a short time.
Then the DJI RTK drones save these inaccuracies as values.
So far, this did not worry me, because these data are included in a compensation by Metashape.
And so far I have been able to really reproduce it.
Images with very poor positioning quality didn't have much influence on the result.

Now I noticed the following:
In the attached image I sorted the approx. 1200 pictures according to calculated accuracy.
I noticed that this deviates in some cases very much from the saved "DJI" accuracy. (See marked line)
If I deselect these images as reference, the result to my GCP is much better.

The stored values within the drone do not always seem to be correct. (Or with a certain time offset.)

Metashape should offer the possibility to take over the calculated accuracy.
At the moment I test this via "reference export" and immediate "reference import" with exchange of the DJI values against the calculated ones.

Do you have any experience so far?

68

General / Re: Processing "dewarped" P4 RTK images....

« on: April 19, 2021, 09:57:35 AM »

Thank you,

The "normal" SHOW IMAGES from menu help also.

69

Feature Requests / Edit thin pointcloud

« on: April 19, 2021, 09:41:22 AM »

Hello.

I often need more flexible processing from the thin point cloud.
You can export it, but you can't import it as a thin point cloud.
Also, you can't crop the thin point cloud "by boundary shape".

Why do I need this?
I have a lot of images and a very "wide" thin point cloud.
But I only want to calculate a certain section of it as a dense point cloud.
The boundary itself can then only be set as a rectangle.
Or I have to select and delete "manually" in the thin point cloud.

I think that if you can crop the dense point cloud according to the boundary shape, this should also be possible with the thin point cloud...!

THX

70

General / Re: Processing "dewarped" P4 RTK images....

« on: April 17, 2021, 04:47:58 PM »

I have seen some things in the FAQ section of DJI.

Frankly, I find it a bit confusing, especially the last section. (BOLD)

Does the Phantom 4 RTK camera digital signal processor (DSP) support distortion correction?

No, you can enable the distortion correction in the camera but the images will be less precise compared to distortion elimination in post-processing software.

How is the Phantom 4 RTK camera calibrated?

Each Phantom 4 RTK camera goes through a calibration process that measures the distortions of the lens, and records the corresponding undistort parameters in XMP DewarpData. When the “distortion correction” is switched off, the original image with distortion will be generated. When the “distortion correction” is switched on, the parameters used to eliminate distortion is not the parameters calibrated, but the design parameters of the camera. When eliminating the distortions, it is not done pixel by pixel.

So the camera internally does NOT use the data stored in the file for distortion correction?

71

General / Re: Processing "dewarped" P4 RTK images....

« on: April 16, 2021, 12:32:07 PM »

It works in part now, but is it right?
I have set the correction values to zero except for the focus.
Then I "fixed" these values. (Picture)

Now it aligns the pictures at least roughly.
After that I can also "optimise the cameras".

I'm not quite sure about the precision of the result yet.

72

General / Processing "dewarped" P4 RTK images....

« on: April 16, 2021, 10:30:19 AM »

Hello.

My software for the DJI P4 RTK drone has a disadvantage.
Photos taken manually are lens corrected by default. (dewarp)
I have also turned this off on 4 from 5 jobs.
I forgot to do it on the last job.

Now I have a chunk of about 1000 "correct undewarped" photos and 100 photos that unfortunately don't fit (dewarped).
The correction parameters are saved in each image and also read out by Metashape.

Can't Metashape take these values to create uncorrected versions for me again or include them in the calculation?

Thank you!

73

General / Technical question about the Rollingshutter correction

« on: April 10, 2021, 12:38:00 PM »

Hello.

I have a question about processing videos.
My "camera" has a hybrid shutter.
The normal photos that I load into Metashape are seen as with a globalshutter.
For videos the rollingshutter method is used.

When importing video frames all Exif data is lost. (Or are not present in the video at all).

Does Metashape not need the exposure times of the photos for the "rolling shutter" correction?
(Or can I improve the correction by adding the exposure times manually?)

Thank you!

Regards

74

General / Re: Automated mission planning

« on: March 03, 2021, 01:38:30 AM »

@3DWinter @MaciekK

For Phantom 4 RTK it is recommended to use DJI Pilot for Android (tested, works fine).

...only if you use the SDK DJI controller.
Otherwise it works NOT. 

75

Bug Reports / Merge & classification problem...

« on: January 27, 2021, 11:57:17 PM »

Hello!

When I merge 2 classified point clouds, unfortunately all classifications are merged to one layer.
I have observed this problem in 1.6.5 & 1.7.

Surely this can be solved?

THX