Topics

1

Python and Java API / Console output and stdout

« on: May 27, 2021, 09:53:14 AM »

Hi!
I am including metashape api (version 1.5) into a larger processing chain, and my goal is to tame the console output a bit, and only log essential values from the alignment/densecloud/orthomosaicing process. Unfortunately this does not work easily, since there is no variable to control the verbosity of the chunk..*() functions. Redirecting stdout (via contextlib or directly) does also not work for some inexplicable reason. Any idea how to deal with this?

2

General / Dealing with Reference Height Inaccuracies while creating a DEM with a Phantom 4

« on: November 21, 2017, 11:44:27 AM »

Hi,
I am trying to align about 1500 Phantom 4 images of a wheat field, shot from about 37 meters above ground. The aligned camera heights show a rather large variance. The image example shows the reference information in green (from image exif), the aligned cameras in red and the optimized cameras in blue. Optimization seems to counteract the large spread in altitude, but it’s still quite bad. I tried to add some constraints to the estimated height with chunk.camera_location_accuracy. I set the camera accuracies to 1m but this does not seem to reduce the spread and the bias at all. Any ideas?

3

General / Altitude Accuracy of Crop Mappings with Phantom 4

« on: April 20, 2017, 03:05:53 PM »

Hi,
I am trying to map the height of a winter wheat field of around 14 hectares with a Phantom 4 (I already know due to the camera the vehicle is not perfectly suited for this, but it's the only one I have). I tried loading the roughly 1500 images into agisoft with python scripts, varying alignment quality and other things like the newly added rolling shutter compensation in the optimization process. Additionally I have 13 markers on the field, measured with topcon hiperv.
The problem is, these markers are all around the field, not inside.
The attached images show camera positions (green: reference from exif, red: position after alignment, blue: position after optimization) and marker positions (green: reference, red: estimate after optimization) as well as the correlation coefficients for latitude, longitude and altitude.

Now I have 2 basic problems, and hopefully someone more experienced than me can add something to this:
1: Latitude and Longitude don't change much after alignment, but altitude changes a lot.  The Variance Increases a lot, even though the UAV had a pretty stable height during the flight. Additionally, a "hanging" effect is clearly visible in the middle of the field. This makes sense since there is no reference data there, but I hoped the rolling shutter compensation would counter this effect. Sadly it did not help much, but in some cases the curvature is reversed. But generally altitude is still all over the place. Any Ideas how to tackle the issue?
2: In some cases correlation of reference and estimated altitudes is alright (like in this case here) but if I up the accuracy, the correlation gets way worse. For instance PS.Accuracy.MediumAccuracy looks good but PS.Accuracy.HighAccuracy results do not? I thought higher alignment accuracy has no downsides (aside from the increased runtime)?

Thanks for your time

4

General / adding non DMS location data to exif of images

« on: March 16, 2017, 04:53:45 PM »

Hi,
I am trying to process around 1500 DJI Phantom 4 images to produce a 3D Model, and as many others I am trying to deal with gps altitude errors and such. I have found a way to replace the altitude with something better (i think), but I would also like to replace the long and lat coordinates with values from my local coordinate system, which is in meters. Sadly this is not possible since  the GPSinfo tags are in this weird format with three fractions for degrees minutes and seconds. Is there a way to add geolocations to the exif info directly without importing camera geometry with csv files?
Thanks!

5

General / Photoscan Peformance on Dell server without GPU

« on: September 25, 2015, 11:25:04 AM »

Hi,
just wanted to know if anybody has got experience running photoscan for aligning and point cloud generation on a server with no GPU. The server is a Dell r520 with 2*6 core Xeons with HT. It also features 96 GB of RAM. It will be running Ubuntu. How will the performance of this system compare to a regular desktop machine with dedicated GPU (like an i7 GTX 980 combo with 16 GB RAM)
Any ideas?

6

Python and Java API / Number of points in the dense_cloud object

« on: September 11, 2015, 10:35:56 AM »

Hi,
I'm trying to write a python script to generate point clouds out of images of plants. I want to log some metadata, and have the following two problems:
1: Is there a convenient way to get the number of points in the dense cloud, similar to how it works in the point_cloud member (len(chunk.point_cloud.points))
2: In the string representation of dense_cloud, a number of points is given like this (<DenseCloud '697114 points'>)
But this number does not match the number given in the TOC (427608 in this example), nor does it match the number of points exported via chunk.exportPoints() (427593 in this example). Why is that?

Thanks in advance