Pix4D Pricing Structures

LutinMalefique · Nov 2, 2017

Just has a quick note, Drone2Map from ESRI is based on Pix4DMapper technology. Unless you have ArcMap in your daily workflow it may not be the most attractive solution (Pix4D have a mounthly subscription, and you can work with QGis to vectorise data for example)

About data, i use to work with a 3DR solo for mapping, someone could share some E90 pictures from a survey mission? i want to process the data to see the quality.

PS : there is also Mavis from http://www.bitmapping.de/. Not tested but seems good and cheaper

Graham · Nov 2, 2017

LutinMalefique said:
About data, i use to work with a 3DR solo for mapping, someone could share some E90 pictures from a survey mission? i want to process the data to see the quality.

I can put some photos on Google drive or drop box. Start a conversation with me via the envelope symbol which should be top right on your webpage if you wish.

Graham · Nov 5, 2017

I downloaded 3Dsurvey Work flow / 3Dsurvey and have been working through the tutorial videos. I loaded some photos from a recent survey and have produced a pointcloud, DSM and orthophoto. I've also done some measurements. All relatively easy given my inexperience. The orthophoto is quite detailed, but I'd like better.

LutinMalefique · Nov 6, 2017

@Graham As i have told you, some calibration should be done, every software has its procedures, I've tried your data with 2 differents softwares (Photoscan an Reality Capture) and i've the same result : a curve similar as fisheye from a GoPro : the pictures are not in a straight line and the shape of the terrain is not realistic, so the orthophoto may not be as accurate as it should.
It can be corrected....photogrammetry softwares are not so "click just one button" result.
View media item 622 View media item 624

Graham · Nov 6, 2017

Interesting, you wouldn't try 3Dsurvey? I've run it on the same data and don't seem to have a curve in the sparse point cloud. It's making a dense reconstruction at the moment, so I can't post a pic to show you.

arruntus · Nov 6, 2017

The images do not have to be in a straight line, neither at the same height nor with the same inclination. The software is in charge of "joining" the images with the correlation of points of the different images and "putting" them in their position.

When an image is taken in a terrain with different elevations the ideal is that in each zone the drone is at the same height to try to achieve a consistent GSD. If this is not the case, the average GSD is much higher. That is why it is so interesting to have the possibility for the drone to monitor the terrain, in order to achieve uniform results. The same thing happens when it comes to inclination, a gust of wind the drone bends and does not take the photo completely perpendicular to the terrain, nothing happens, the software takes care of processing it taking into account the azimuth inclination, and so on.

In fact, you can work without GPS coordinates, it costs you more, but you still get it. What is not achieved in this way is to have a model neither scaled nor georeferenced but can be fixed later with the GCPs and checkpoints.

LutinMalefique · Nov 6, 2017

arruntus said:
The images do not have to be in a straight line, neither at the same height nor with the same inclination. The software is in charge of "joining" the images with the correlation of points of the different images and "putting" them in their position.

When an image is taken in a terrain with different elevations the ideal is that in each zone the drone is at the same height to try to achieve a consistent GSD. If this is not the case, the average GSD is much higher. That is why it is so interesting to have the possibility for the drone to monitor the terrain, in order to achieve uniform results. The same thing happens when it comes to inclination, a gust of wind the drone bends and does not take the photo completely perpendicular to the terrain, nothing happens, the software takes care of processing it taking into account the azimuth inclination, and so on.

In fact, you can work without GPS coordinates, it costs you more, but you still get it. What is not achieved in this way is to have a model neither scaled nor georeferenced but can be fixed later with the GCPs and checkpoints.

but if you do a survey mission, all the pictures are taken following straight lines at same height.
If the terrain has different elevations, the drone should keep the same height to have the same GSD but that fonction is not a standard. In my memory, you can do that with Mission Planner (elevation is from Google Earth) and some versions of Arducopter.
But in the case of Graham's data he gave me access to, it's lens distortion, i pretty sure of that. It's not a lens quality issue, it's just how software interpret data without any other information.

LutinMalefique · Nov 6, 2017

@Graham, i will take a look at 3DSurvey.

javier Olalla González · Nov 6, 2017

In photoscan, in the camera calibration tab, you have to select a fish-eye camera. Then refresh and the model is corrected immediately.

arruntus · Nov 6, 2017

The problem here is that it's not supposed to be a fish-eye camera.

Something strange has happened with those images or with the taking of them or something. See if we can find out what it could have been.

By the way, @javier Olalla González we have two open conversations pending. I don't know if you've denied the conversations and not read me.

javier Olalla González · Nov 6, 2017

Hola, buenas noches. Te respondí por e-mail. No lo recibiste?

arruntus · Nov 6, 2017

javier Olalla González said:
Hola, buenas noches. Te respondí por e-mail. No lo recibiste?

I have not received anything, the conversations between users are in the title bar next to your name, a letter envelope. Don't you have anything in there?

ArltTech · Nov 8, 2017

Here is a quick test mission using a trial version of DroneDeploy. In Vimeo settings be sure to set to 1080p. H520 E90 GSD set to 1.0 cm/pixel survey mode in DataPilot.

Graham · Nov 8, 2017

Arruntus, LutinMalefique and I have been cooperating in an assessment of the H520 and E90 survey facility. The attached survey was planned using Datapilot on a laptop in a matter of minutes, transferred to the H520 and flown without any problems. A photo of the DataPilot survey plan, Microsoft ICE stitch of the photographs and report made using pix4d are attached.

javier Olalla González · Nov 9, 2017

Good job guys. Interesting data is obtained from the report. Despite not having taken GCP (Ground Control Points ) there are several parameters that indicate that it is a good model. A GSD of 0.92 cm / pix.
A difference of 1.95% between the initial values of the camera and the final values optimized by PIX4D. A small retro projection error (0.330 pix). And all the georeferenced photos with errors between - 3m and 3 m.

javier Olalla González · Nov 9, 2017

I can not understand why there is an 8% of uncalibrated cameras.

arruntus · Nov 9, 2017

Yeah, it's kind of weird. In the middle of it, a few images fail. I've given it a lot of thought and I don't know why.

It's a two-flight shot, it only occurs to me that it's the area where the flights are linked but it covers two flight lines, if it's just one..... that doesn't make sense either.

bluelight.support · Nov 9, 2017

Graham said:
Arruntus, LutinMalefique and I have been cooperating in an assessment of the H520 and E90 survey facility. The attached survey was planned using Datapilot on a laptop in a matter of minutes, transferred to the H520 and flown without any problems. A photo of the DataPilot survey plan, Microsoft ICE stitch of the photographs and report made using pix4d are attached.

Graham - can I try that dataset in Photoscan Pro - I have been having a few issues with the other one!

Graham · Nov 10, 2017

bluelight.support said:
Graham - can I try that dataset in Photoscan Pro - I have been having a few issues with the other one!

I'd be pleased to. Though we now have something better!

Just in case somebody with the authority to make amendments to the DataPilot software is reading this, in the programming box on the right, meters should be spelt metres. Annoys me every time!