Phantom 4 and Pix4d

Yes, these were all the changes I made in that project.

There are two camera models in the database and for image dimensios 4000x3000 I had I used this camera.

Hope everything goes well !!!

 

It worked well thanks. 0.47% relative difference. Where did you read the 2.9cm RMS error from?

Perfect, you have to put GCPs and you will find it in the report !!!

Much appreciated many thanks Ilias, you really helped me a lot.

Glad to help !!!

@ Ilias Pontikas and @Dirk Grobler

Mobile App: Pix4d Capture (android)

Desktop App: Pix4D trial

UAV: DJI Phantom 4

Hi guys, I am facing the same problems you debated above. But in my case, whatever changes I make in the camera options I still have above 90% relative difference. I do not know what I am doing wrong. I flew with my P4  an area 150x180 with a simple grid mission at 50m alt. and it took 170 pictures with the camera facing straight downwards. The only issue I had was after the mission ended where I have interrupted the synchronization process. Either way I have used the pictures from the P4 SD card.

The tie points in my project are flat, there is no morphological aspect to the whole project and the mapped area is across a sloped terrain with sparse houses. 

I haven’t used GCPs for this project.

Do you have other recommendations?

Thanks in advance,

Cris

Hi Cris

Before getting into detail just confirm the following:

  1. The synchronization will not affect the project. As long as you have the photos you’re good to go.

  2. What was the overlap? Should not be below 75%. If monotonous terrain with very little visible features at least 80%.

  3. Did you select the correct camera? Must be FC330_3.6_4000x3000.

4 At “Shutter Model” you must select “Rolling Shutter” for the Phantom 4.

  1. Use at least Pix4D version 3.0.7. Version 2.2.25 had a bug in the rolling shutter algorithm.

 

Let me know if the problem persist after fixing the above.

Regards

Dirk

Hello Dirk,

Thank you for the prompt response!

  1.  Yes, I assumed this also. I wanted to start a project from scratch and this was done using the pictures from the SD card from UAV.

  2. The overlap used was 80%!

  3. Yes, this was the camera used for processing.

  4. I have used Rolling shutter but the initial processing failed and so I could not do anything Unfortunately I have not written down the error code. I will give another try and I’ll get back to you with the code.

  5. I have used the version 3.0.13 from the pix4d website.

Additionally I have used two different PC with different configurations but capable of handling this job (i7/i5, 16Gb RAM, 780GTX NVIDIA, win 10)

Today I have updated the latest firmware released two days aggo from DJI. Maybe this will solve my problem. I have another set of pictures and I will give another try and see what happens. I will let you know ASAP the results.

Any other recommendations?

Regards

Do you think the windy conditions have to do with my errors? I Had between 30-27 Km/h wind speed.

Hi Cris

Windy conditions will definitely affect quality of processing in my opinion. Just think about it. If the drone has to swing for example 30 degrees to maintain position against strong wind, imagine the difference in scale from one side of the photo to the other.

I do not fly in winds higher than 20 to 22 km/h MAX. Even then I can see the Phantom 4’s legs on some of the photos as the drone negotiate gusts of wind in excess of that. The higher up the stronger the wind.

I suggest submit a request to Pix4D and get an official response on this.

You should ideally run the complete job on one PC. If you ran the complete project on both PC’s, do both give you bad camera optimization results?

Regards

Dirk

Hi Dirk!

Today I made 5 flights with my P4 with pix4dCapture. Just before these flights I got a warning that there is a new firmware for the P4 so I have updated from July’s firmware 01.02.0503 to November’s firmware 01.02.0602. I made these updates and went in the field. I got lots of pics, flew with simple grids and also double grids. So, busy day! 

Now, at home, I made a initial processing for one of today’s flight and guess what? 9.35% relative difference! with rolling shutter and I used the “Estimate from EXIF” for the parameters of the camera and I get 9% over 90%. Huge difference!. The weather here is windy and the chances are very slim to drop so I will keep on practice until I get better results. 

Is the 9% RD (relative difference) OK for a non GCP flight?

I am very curious for the rest of the today’s projects if the RD will stay in 9% range. 

The previous project that generated this discussions clearly had some problems. Either the firmware of the P4 either something else… I will send to support the report with those errors. regarding the two PCs running the same processes: I did not got to finish the processing on either of them. I got some error codes within the Log Output and the tie points, like I have described before, were flat.

I will get back with brief details about the rest of the today’s results.

Sorry for the long description,

Regards,

Hi Cris

You clearly made some good progress. Very liberating feeling I know.

RD must be below 5% so you’re very close. A bit more playing around and you’re there!!

One very important tip. After running the project for the first time, go back to “Select Camera Model” and click “Edit”. Then click “Load Optimized Parameters”. That uses the first round’s calculated parameters to run a second round where you should get major improvement to the RD figure.

Let me know if this did the trick.

 

Regards

Dirk

Thank you Dirk for all your help. The rest of the other flights are a mess though. Seems that the wind is to blame here. Now I have 7km/h outside and I will get in the field again.

Another thing is that a friend of mine lend me his Ipad and there is an option in the software where i can tell the UAV to stop and make picture within the mission. I think this will improve the quality but this option is not available on Android. Very curious!

So, you tell me that I should make an initial processing with no tweaks to the camera and then repeat it with the rollng shutter and optimize camera from EXIF? I will give it a go and see what happens.

Thank you so much Dirk! Safe flights and keep in touch.

@ Adam Orens

I do not know about their future plans but the app works great! The processing is a bit tricky but read the conversations in this thread and you will be good to go. Also I have noticed a small difference in the Capture apps from Android to IOS (description above) which is surprising.

Cheers!

 

Hi Cris

First round should be camera as selected on your previous projects with EXIF and including rolling shutter. Only difference with second round is you load optimized values.

Allow me… Dear Dirk,

Your suggestions worked out perfectly! I got 3% RD on one project and on the other 0.8% RD !!! For the second one where I got 0.8% was without LOP “Load Optimized Parameters”. By loading LOP the error increases all the way to 8%. No worries, is good to know that option.

I also had a perfect weather today, 5km/h wind and sunny. Plus I have mapped with the apple device with the option to stop and take pictures along the track (I am sure that this option helped a lot but the battery…).

I still have problems with the calibration of some pictures (ex. from 130 pics only 90 calibrated). Any advice on this topic?

Thank you again Dirk and I will have a beer in your name :slight_smile:

Cheers,

Cris

Hello everyone,

In my opinion the major problem of P4 is the rolling shutter, you can’t achieve easily good results, you must have ideal conditions or stop and take the photos. When it moves it should have the same speed that the algorithm will use to make the corrections, with windy conditions that’s not easy. I have achieved accurate results without winds, with a big overlap, by taking photos in two directions and using a lot of GCPs for better external orientation.

The external calibration of the cameras has to do mostly with the pattern in the photos, there must be discrete points for the relative orientation to take place, otherwise for few photos it could be done manually. 

After a lot of reading, I landed in this thread and felt it’s the place for me…I’ve been facing an issue with my 1-month old P4P while using Capture!

After plotting a grid, either for 2d or 3d surface models, the drone takes off, reaches the start point (sometimes with noticeable delays), starts acquiring images, goes along the first row, returns along the second AND STOPS!!!

I spent 4 hours in the field today, had to drive 3 hours to and from the place, emptied two batteries and came out with nothing…

I understand that Pix4D is a great tool and I have to admit that I had great results ONCE! But it was only once!!!

Working with latest Pix4D Capture, DJI GO 4 updates and Google Nexus 5 (Android).

Thanks in advance…

 

P.S. Somebody mentioned greenish or/and blockish images in DJI GO 4 app…Check this out and weep…I did…spent 3-4 days for a solution and this kid’s trick worked for me…I was enraged…

https://www.youtube.com/watch?v=NxqCWnPpkRM

 

 

Hi Triantafyllos,

I’m sorry to hear about this experience. I would first suggest to check the recommendations of this article:
https://support.pix4d.com/hc/en-us/articles/115004119343

Let me know if they do not help so I will investigate further. Also I would recommend to try with another Android device than Nexus.