Recap photo GCP-Magna Bogotá Issue

[Anonymous]

Hi! I have installed recap photo with my student membership and I have realized that now it eats cloud credits. But that is not the problem, which is that the new software doesn't work properly. I processed 98 aerial photos taken by DJI Phantom with 3GCPs surveyed by RTK. The interface is simple and friendly, same as ReMake, and is too easy to select the GCPs. The photos were taken in WGS84 (Deg, Min, Sec), but the LLC WGS84 format is only Deg. ReCap Photo recognizes the GCPs' coordinates in Magna-Sirgas Bogotá Zone, entering it by X, Y, Z (not N, E, Elev). When we select FILTERING BY GPS option, the software shows the images in which the selected GCP appears, filtering by GPS coordinates (Phantom photos). There is the first issue. Recap Photo showed me pictures where the GCPs doesn't appear, and showed the same images selecting 2 different GCPs. Why??? I verified that I have selected (right) the GCPs according to the surveying data and entered the rigth coordinates and then, all was done to process. When I download the generated data and waste 12 cloud credits, I realized that the 3d model (point cloud RCP), the mesh and the Tiff orthomosaic were wrong located. The points were displaced around 5m at X and Y and around 12m up the GCPs elevation. Why???? It is assumed that control points help to force the coordinates to the real position, but I am so disappointed. I wish to make an extensive research about it, but now I got the limiting of the cloud credits, I waste almost all of them. Please, if you are trying to sell us credits verify first that your software REALLY WORKS, you are very far from software like Pix4D, which has a clean performance. I love Autodesk and all your software, but Recap Photo was a really disappointment! I hope you can fix that problem a giving us a fully functional software that we can paid with pleasure. Please see the attached images. Thanks for your help!

[Anonymous]

Hi Andres, I am from Bogota too but I live in Florida USA and would like to contact you regarding GCPs , please let me have you e-mail so I can contact you.

Mine is :

[email protected]

Thank you so much.

[karthik.nathan]

Hi,

Thank you for the feedback. If you don't mind sharing your dataset (photos & output files) with us through A360 Drive (https://a360.autodesk.com/drive/) or any other platform, that will help us see what's going on. My email address is [email protected].

Thank you

-Karthik

[anil_mistry]

Hi @Anonymous,

I’m just following up to see if you were able to gather the information @karthik.nathan had requested yesterday to assist him in troubleshoot your issue.

Thank you and have a great day!

[Anonymous]

Ok. I will share the project with you. Thanks for your help!

[Anonymous]

Thanks for your interest. I´ve shared the project with @karthik.nathan trying to find a solution. Best regards!

[karthik.nathan · Accepted solution]

Hi Andres,

Thank you for sharing the dataset. We found a couple of issues in the gcp data you entered. Adding GCPs and selecting them on images through ReCap Photo generates a *.gcp file with this info that is used by the engine. This *.gcp file can be found in the A360 folder you shared with us. Here are the contents of that file ...

<?xml version="1.0" encoding="UTF-8"?>
<surveydata coordinatesystem="LL84" description="MAGNA-SIRGAS / Colombia Bogota zone" epsgcode="3116">
<markers>
<marker id="0" name="1">
<images>
<image name="DJI_0027.JPG" xpixel="1236" ypixel="2049"/>
<image name="DJI_0029.JPG" xpixel="1177" ypixel="1216"/>
<image name="DJI_0030.JPG" xpixel="1150" ypixel="821"/>
<image name="DJI_0031.JPG" xpixel="1115" ypixel="417"/>
<image name="DJI_0045.JPG" xpixel="2549" ypixel="561"/>
</images>
<gcp x="-yyyyyyyyyyyyyyyyyy" y="yyyyyyyyyyyyyyyyyy" z="yyyyyyyyyyyyyyyyyy" checkpoint="true"/>
</marker>
<marker id="1" name="2">
<images>
<image name="DJI_0016.JPG" xpixel="527" ypixel="2037"/>
<image name="DJI_0022.JPG" xpixel="1762" ypixel="2031"/>
<image name="DJI_0024.JPG" xpixel="1714" ypixel="1241"/>
<image name="DJI_0025.JPG" xpixel="1678" ypixel="823"/>
<image name="DJI_0026.JPG" xpixel="1650" ypixel="436"/>
</images>
<gcp x="xxxxxxxxxxxxxxxxxxx" y="xxxxxxxxxxxxxxxxxxx" z="xxxxxxxxxxxxxxxxxxx" checkpoint="true"/>
</marker>
<marker id="2" name="3">
<images>
<image name="DJI_0063.JPG" xpixel="2624" ypixel="2406"/>
<image name="DJI_0064.JPG" xpixel="2605" ypixel="1971"/>
<image name="DJI_0065.JPG" xpixel="2586" ypixel="1510"/>
<image name="DJI_0066.JPG" xpixel="2565" ypixel="1084"/>
<image name="DJI_0068.JPG" xpixel="2511" ypixel="282"/>
</images>
<gcp x="xxxxxxxxxxxxxxxxxxx" y="xxxxxxxxxxxxxxxxxxx" z="xxxxxxxxxxxxxxxxxxx" checkpoint="true"/>
</marker>
</markers>
</surveydata>

1) It seems all the points / coordinates entered were entered as checkpoints and not ground control points. Did you select the 'Checkpoint' checkbox for every coordinate entered by any chance? Check points don't influence the reconstruction and are only used for validation / verification at the end. This likely explains the bigger than expected deviation that you observed.Ground control points on the other hand greatly influence the reconstruction greatly and will result in a much smaller deviation (if any).

2) We tried to set 'checkpoint=false' in the gcp file for all points and re-run the reconstruction but noticed another issue. It seems the last two points have exactly the same coordinates but have been marked on different images at different locations. Was it perhaps a copy paste error when entering the coordinates in ReCap Photo?

If you can try again with these two issues corrected, hopefully that will give you the result you expected. Or if you can give us the correct coordinates for either point #2 or #3, we can re-run the reconstruction and share the results with you.

Hope this helps.

-Karthik

[karthik.nathan]

Hi Andres,

One more thing ... the coordinates being the same for two points probably also explains why the filter by GPS option was displaying the same set of images for both points. We will look into how we can improve the workflow and do some validation on the software side to prevent such scenarios.

-karthik

[Anonymous]

Thank you, very much, I really appreciate your efforts!

1. Yes, I entered the 3 points like checkpoints, because i Guess that is the workflow. Now I've understood, No check points. Ok. Thanks.

2. The second point is the real issue. I made copy-paste of the coordinate table:

PUNTO	COORDENADA ESTE	COORDENADA NORTE	COTA
1	1034742.474	1266132.783	86.4264
2	1034702.620	1266191.810	83.7744
3	1034702.445	1266096.529	86.7394

There isn't copy-paste error, the coordinates are right, but the software reconigzes the second and the third point as one, so much so that, when I was selecting the GCPs with GPS filter enabled, the software showed the same images for the 2 and 3 point.

I will try again without the marking of checkpoints. Maybe that the reason for the issue. 

Thank you very much again.

[Anonymous]

Well! I've send again the project without marking checkpoints. No one. The software have done a great job, and it is confirmed: The issue was about the checkpoints, I had marked all of 3 GCPs as checkpoints and this cause (probably) that the coordinates of the 2 and 3 GCPs were duplicated. I suggest to look over that Chekpoints and their funcionality, I don't understand very well, what are they for?

Thank you very much @karthik.nathan for your review and information. I will continue in my research of recap photo, but the worse news for me: I haven't any credit!!!!

[murali.p]

Hi,

Check points are used for quality assessment of the calculated point versus the ground truth. This information is available in the report. 

Typically, 5 GCPs at the minimum are recommended and you can use additional GCPs as check points.

Regards,

Murali

[Anonymous]

Hello
I have the same problem, I have measured 5 GCP in coordinate system Sweref 99 1800 when I insert points in rcp
do my pictures appear right but my points in the middle of the sea, what am I doing wrong?

[daniel_musick]

I answered in your other thread as well.

You forgot the correct decimal points and switched the values for x and y....