Support Website Contact Support Blog

Different RMS errors from same data

 Hi!

I did some processing test with pix4d a while ago and I was studying again all the data and outputs from it. The data was from a corridor map data from sensefly. I noticed that the quality report from both proccess are different (sensefly RMS error is just mm and I got cm). What could be the reason for this different accuracy in absolute geolocation data in both projects? Could it be the set of parameters, or maybe is according the capability of process of the computers?

Here are both quality reports, from sensefly and from my processing.

https://drive.google.com/open?id=1od4xrckercu9ly0-agu__lYy63ChIs-B

https://drive.google.com/open?id=1RTYezAaod_-r8spaSrkv_LfsudcjCQ5K

 

Best Regards 

 

Hey Victor,

If you are changing the parameters it may definitely have an affect on the results and computer specs may also change the results as well. I had projects before I processed it twice in the same computer using the exactly the same settings and I got very slightly different results in the quality report.

Hey Selim, thanks for your reply. Could you take a look at both reports in the links I left in my comment? I think I used all the same parameters but still I can’t find where could it be the difference.

Hey Victor,

I was able to take a quick look at your quality reports. I have noticed one project was processed in Linux using the Enterprise licence version 3.2.23 and the other one was processed using the Pix4D Mapper Pro version 4.1.22. There is a significant difference there, there is almost a year between the software versions and they are in different platforms. Also camera optimization between initial and optimized internal camera parameters is different 0.27 vs  0.81. Those are probably the reasons why there may be a discrepancy. That being said however both of those project have RMS errors in cm’s not in mm’s. First project RMS X is 0.003727 (0.37 cm’s) second project RMS X is 0.010922 (1.09 cm’s). In my opinion they have pretty similar RMS X, Y and Z errors even though software versions and platforms are significantly different. 

Hey Selim, thanks for taking time looking at the reports. Funny thing is the one made one year before have better final accuracy. I know the difference beetween them might be very small, but I am looking results around 0.37 cm with rtk activated in other projects. Thats why i am trying to stay around that number. But my process got 1.09 cm, and I didnt get why the difference. You point out the camera parameters, i didnt notice that. How can I improve for re processing?

Hey Victor,
As long as the camera optimization between initial and optimized internal camera parameters are less than 5% it is accurate. If you want to tweak with the internal camera parameters settings, it is under Calibration/Calibration Method/internal parameters optimization change that to all prior. Before changing those settings I would however recommend maybe changing the image scale to 2 instead of the default image scale 1. You may be able to improve the accuracy a little bit. Moreover as long as your accuracy is 1-2 GSD vertically and 1-3 GSD horizontally, your project is pretty accurate.

Selim,

Thanks for taking the time to explain to me all this information. I will take all that into account for next time I process data.

Best regards

Hey Victor,

No worries.