The only offset that I apply is the height between antenna and camera. It is about 21cm.
I am trying to find a formula to address the tilt of the drone. I tape measured the offset to be about 3cm while the drone is tilted. It is doable but it is supposed that the drone has constant speed.
About the shutter delay how can this be corrected?
I compared a p4rtk dataset a friend sent me and the camera errors were 2-3cm on a single grid mission. Also some videos on YouTube show the camera errors about 2-3cm.
I suppose that p4rtk is way more sophisticated in terms of accuracy.
Hi @vgo195,
Havenāt you checked the standard deviations in the events.pos file before importing images into Metashape? You can send me the events.pos file along with the logs from the base and the rover via PM or via support@emlid.com. Iāll check the geotagging accuracy, and try to post-process logs.
I assume that this error you obtained may be related to imagesā referencing accuracy. Apart from image coordinates precision, a droneās tilt, wind, and photoās distortion may affect it. Using GCPs can also help mitigate their impact and improve the total accuracy of the orthomosaic map.
I also suppose that the manual photosā accuracy is higher due to the less number of photos and less impact of factors above.
Maybe you know what value should I use for the shutter capture lag in toposetter?
+0.03sec or -0.03sec?
The tag is later than the image so you need to backup it up. This sensor is reported to have a 30ms delay. It depends on your workflow and your software capabilities but this gets us close enough that in most situations all that needs to be corrected after processing is the vertical adjustment from ellipsoid height to geometric elevations. In our experience you canāt just calc the geoid separation for a site with the NGS calculator and go. Construction site benchmarks rarely match vertically within the 0.10ft tolerance so we have to adjust our vertical control (localize) to agree. Itās usually within 0.20ft so it is close and can catch you by surprise.
So it is -0.03 sec.
In my country we use a tool provided by the cadastre agency to convert the Ell height to Orthometric so this is solved.
My problem is that I donāt see camera errors after alignment within 5cm. I did another flight and also had 10-20cm camera errors. Maybe I am doing something wrong in metashape workflow.
Thanks anyway!
do we have any person from metatec on this forum? i am interested with this productā¦
Hi @vgo195,
Iāve checked the logs and havenāt noted any suspicious in coordinates accuracy. I also see that accuracy of images coordinates taken during the single grid mission and stopānāgo approach is the same. So, the other factors like droneās tilt, wind, photo distortion may affect the final post-processing accuracy.
Thanks, I agree that these factors play major role to the final accuracy of the model.
Another question, maybe 10hz is not enough for a drone that fly at 4-5m/s? Does M2 support more than 10hz?
Hi @vgo195,
10 Hz should be enough for drone mapping. And itās the highest rate that Reach M2 supports.
Weāre getting 1.5cm H 2.5-3cm V with 5Hz all day. The CORS in our area only does up to 5Hz so it is what it is.
Thanks for the replies.
Another test with āstop ānā goā method gave me good results within 3cm.
I am evaluating toposetter that seems more sophisticated than emlid studio. It takes account various parameters like drone tilt, shutter lag and can use XYZ offsets.
I am not underestimate ES, it is wonderful (and free) app but it seems that it doesnāt fit in the metattec ppk kit workflow.
Maybe someone from metattec should comment on these!
Blockquote
They work in CIAā¦they are invisibleš
Hi
You could also try the UAV - PPK application, Keep the flow of the metattec PPK KIT. It is a plugin for Emlid Studio. You have to enter the .pos and eventpos file generated by ES. The application corrects the delay of the photosensor, recalculates the droneās trajectory and finally generates a .csv with the precise coordinates of each image.
Hi,
I see in your video you apply 0.3sec offset. Is this correct? This is 300ms and other users say that the shutter delay is 30ms. Which one is the correct value for mettatec ppk kit 0.3sec or 0.03sec?
Hi
After several tests, it was concluded that the delay of the Metta KIT photosensor, like the vast majority, is 300 ms (0.3 seconds).
Correcting this delay and aligning the photos in metashape or Pix4d, precision of 0.03 meters is achieved.
Thanks for the replyā¦
I donāt know but 300ms seems too much, maybe someone from mettatec can comment on this.
Hi, I was chatting with Metta Tech Support. The factory delay of the Photosensor is 30 ms, but it must be taken into account that there is an additional delay for capturing the photograph and turning off - turning on the front LED of the Phantom 4 pro.
The default option (KIT PPK Phantom 4 pro ) in UAV -PPK contemplates the total delay of the workflow.
Nice infoš
For me 300ms doesnāt work. I did some tests with toposetter and I didnāt like the results.
I did some tests using the default workflow with Emlid Studio (no shutter speed correction) and I came to these results:
- The accuracy (compared to gcpās) is within 3cm only if āstop 'n captureā method is used and no gcpās are taken account in metashape.
- If you fly a standard mission (single or double grid) then the accuracy is within 5-10cm if no gcpās are used.
- To be able to have cm level accuracy in your project at least 3 GCPās must be used in metashape. The good thing is that with the minimum 3 gcpās the results in the control points are within 2-3 cm all over the site. I did 2 data sets and the results were the same. I want to do these tests in a larger project to see if the distribution of gcpās plays major role in the accuracy of the control points.
These results were obtained by using 10-20cm accuracy value for cameras and the default 0.005 accuracy for the gcpās.
Overall I am happy with the results but still have to do more accuracy testsā¦!
Hi, I am very interested in your assessment on the issue with Mettatec kit and P4P vs metashape. I am experiencing the same exact issues with accuracies in Mettashape. I made a lot of tries to create a workflow using Emlid studio but nothing is certain. Firstly I had issues with extra events created by the blink of the p4p without capturing a photo. Fortunately I used a lot of flight plan apps and figure out that in almost all my flight just before my first photo I am getting an āemptyā event. I addressed that by hacking emlid studio and inserting an extra photo in the start of my dataset. After that I imported the geotagged images ( without the first wrong one) in MS and I got the same exact issue with vgo195 the accuracies are going crazy both in cameras and gcps. I tried addressing the delay issue but to no avail. The most strange issue was that today I tried a trial version of Redtoolbox because their workflow seems nice but I came across to this strange issue. I solved the same flight data on both Emlid studio and redtoolbox and on Emlid studio using the forward mode I am getting mainly fixed solutions in the red toolbox im getting 95% float! This is mind boggling!!! I am using the M+ receiver with the latest mettatec kit on my P4P.