Topics

1

Bug Reports / Convert whether K4 is enabled for the InPHO project

« on: October 30, 2021, 07:55:56 PM »

This issue has been raised for a long time, PhotoScan conversion to inpho project distortion coefficient is wrong.

It was not taken seriously by official engineers

Why is that?

I further verified this problem.

If it is an aerial camera with small distortion coefficient, At this time, K4 distortion coefficient will not be introduced.PhotoScan conversion results work normally. For example, leica, Feisi and other professional equipment.

However, PhotoScan will automatically reference the K4 distortion coefficient for devices with large distortion coefficient, and the result is wrong.

This has been a problem for years.

2

Feature Requests / No quasi-geoid was introduced when reading pos from exif

« on: July 23, 2021, 09:37:42 PM »

The height read from EXIF is ellipsoidal height, which needs to be converted to normal height by introducing EGM. At present, there is no option to enable quasi-geoid in the fully automatic processing of the software. This is a very meaningful function.

3

Feature Requests / Cannot be adjusted to POS

« on: July 23, 2021, 09:33:15 PM »

Usually when optimizing the camera, parameters such as f cx cy will be checked.

If the control point and the Z of POS are not in the same coordinate system, it will cause a lot of trouble. You will find that f is optimized very small.

If you uncheck f, you will find that there is no optimization at all.

This reminds me of inpho, which spreads the residual of the adjustment into the x y z of POS, which effectively avoids the problem of control points and POS elevation z, which are not in the same coordinate system.

4

Bug Reports / Ellipsoid conversion, elevation Z not calculated

« on: July 23, 2021, 09:16:32 PM »

reproduction process,
a calculates the seven-parameter Bursa model
b selects the xyz coordinate system in ms,camera selects wgs84 ,input the seven-parameter
c converts pos to xyz and to check pos

There is no change in z, and z will change in the software that calculates the seven parameters.

Is it possible to add a matching seven-parameter calculation function and save the results as a prj file?

5

Bug Reports / Exported inpho project with incorrect distortion parameters

« on: July 23, 2021, 09:07:17 PM »

This problem has existed for a long time, usually we use distortion-free images, which have not been paid attention to.

Reproduce process

a export inpho prj
b open with inpho, import control point
c in the prickle interface, you will find that the image is deformed.

6

Bug Reports / Import inpho camera POS and distortion coefficient

« on: July 23, 2021, 09:01:13 PM »

I found one problems for POS

a The records in the pos and inpho PRJ imports are inconsistent, and the first line is omitted.

I found two problems for distortion coefficient:
a·Check camera parameters dialog box, all parameters are N/A
b.Multiple cameras or one camera has several sets of parameters (inpho self-calibration)，Unable to recognize。


and 1.7.4 - Pre，There is a problem with the distortion coefficient conversion, which only needs K1-K3 P1-P2.

7

General / PhotoScan AT Conversion to INPHO Tools 【Public】

« on: July 23, 2020, 09:13:09 AM »

I wrote a program to convert PhotoScan AT results into an INPHO project.

The only difference between the public version and my private version is that TiePoint and control point cannot be imported.

My approach has the following advantages:

1, support distortion(k1\k2\k3\p1\p2) transformation, you do not need to export undistortion image.
2. OPK conversion is supported. You can use the Match-AT initialization EO function.

Hope to improve your work efficiency.

Meng Wang

8

Bug Reports / "B1" Not involved in the operation

« on: May 20, 2020, 12:07:22 AM »

I hardly ever use B1.

for DJI 4 RTK

f = fy
k1 = k1
k2 = k2
k3 = k3
p1 = p2
p2 = p1
B1 =  fx / fy - 1, so I've defined B1.

After the photo is aligned, B1 doesn't change at all. I didn't fix B1, which I think is strange.
Usually, self-check will cause the parameters to change, but B1 has not changed.

9

Bug Reports / calibration import from DJI XMP meta data

« on: March 01, 2020, 09:13:03 PM »

My English is generated by translation software.Please forgive me.

3664.650000000000,3658.030000000000,-10.810000000000,22.730000000000,-0.263279000000,0.112381000000,0.001254550000,-0.000833425000,-0.039363700000

Fx Fy Cx Cy  K1 K2   P1 P2  K3   This one Consistent with Pix4D camera model，so in PhotoScan:

pix size = 0.0024

f=(fx+fy)/2 // or  f = fy   and  b1 = fx-fy
cx = cx
cy = cy
K1 = K1
K2 = K2
K3 = K3
P1 = P2
P2 = P1

My opinion can be verified:

Bentley CC is compatible with the PhotoScan camera model, and only self-calibration is used to observe the self-checking results of CC P1 P2.

10

General / The CAMERA CALIBRATION parameter conversion is inconsistent

« on: September 20, 2019, 11:51:47 AM »

Hello,

The parameters of PhotoScan are attached to the report in Australis new 8.33 version,

I observed that the result of the transformation is completely different from the result of the transformation by PhotoScan.

Excuse me, can this difference produce serious consequence?

thanks

11

General / Does keypoint caching really work?

« on: December 27, 2018, 01:20:46 AM »

Keep the key point, do not reset, will activate the cache.

However, cloning a block or merging multiple blocks still requires reextracting the key points.

I really want to be able to cache based on the photo path or file HASH as the identity. (number of consistent points)

I think it helps increase efficiency.

Feasible?

12

General / about "Bad memory allocation‘

« on: December 23, 2018, 05:03:19 PM »

“Bad memory allocation”，It's a long-term problem。

The current of the solution, can not be in a block of tens of thousands of photos adjustment.You cannot adjust（Task distribution has been activated, Final CPU adjustment） >10000 photos in one block; otherwise, this problem is often triggered.

* The current recommendation is < 10000 photos
** Of course, you can do this > 10000 if you reduce the number of key points.
*** If there is a large amount of memory (I have 256G), I have tried > 20,000 photos and > 40,000 photos without causing bad memory allocation error. However, in the end, the CPU adjustment week did not end and I had to give up.
**** Even if you have a large memory , > 13,000 photos, high risk of triggering an error. In addition, you may also trigger an upper limit on key points.


At the end of this year, some new software appeared, which is not perfect, but they can handle the adjustment of  >10w and <15w photos.

In the future, can we provide new solutions for bad memory allocation?

13

General / Different export formats, why XYZ is different?

« on: December 12, 2018, 06:42:19 PM »

sorry ，It's all my fault 。

Please help me delete this post, thank you.

14

Feature Requests / new projects with images list？

« on: December 11, 2018, 08:29:36 PM »

Hello Sir,
 
I have a list of image files,but there are other files in the image folder,At the same time, different projects need the same image file.moving and copying image files is not a good idea.


image lists :

x:\123\1\1\1.jpg
x:\456\1\2\1.jpg
x:\ABC\3\1\100.jpg

Typically, the subdirectory is a photo of the same physical camera.

15

Python and Java API / Please advise how to keep only photos with POS.

« on: May 24, 2018, 11:19:54 AM »

I have a 5-camera multi-lens system, and only orthographic with POS. For large projects, I hope to split into multiple blocks for photo alignment.

If I delete the in main gui with blue dots, the other four photos without POS are not deleted.

I think we can have a script that will delete all the extra photos through the file name rule.

The file name rule is as follows (1):
File names are unique throughout the project
1 x001 --- Ortho, with POS
1 q001
1 h001
1 z001
1 y001
……

2 x001 --- Ortho, with POS
2 q001
2 h001
2 z001
2 y001
……

If it doesn't exist 2x001 Delete the 2q001 2h001 2z001 2y 001.

The file name rule is as follows (2):
The file names are the same, but are stored in different 5 folders.
001 --- Ortho, with POS
001
001
001
001
……
002 --- Ortho, with POS
002
002
002
002
Can you give me a hand?