Agisoft Metashape
Agisoft Metashape => General => Topic started by: piste on October 01, 2012, 12:24:28 PM
-
Good Morning.
Do I have some photos taken during a flight Lidar.
And there is also the point cloud Lidar and orthophotos.
I want to test the goodness of the point cloud extracted from the geometric model obtained from Photoscan with respect to Lidar point cloud.
I noticed that it gets a good planimetric accuracy but it's wrong a few meters in altitude.
I do not have the file available for calibration of the camera (Hasselblad CF-39).
The model has been georeferenced by taking the z-coordinate of the point cloud Lidar x and y coordinates from orthophotos.
I am looking for some advice to get better results.
I will add some screenshots to better understand the problem.
The purpose of this test is to be able to integrate in the future the point cloud we get through a terrestrial laser scanner with a cloud can be obtained through a photogrammetric analysis of pictures taken by UAV or aircraft.
Thank you very much.
David
P.S.: Is it always necessary to insert x, y, z coordinates of camera positions in pre-alignment stage?
If I have only marker coordinates, are they usable for georeferencing model with more accuracy?
-
Have a look at the two datasets in cloudcompare. That should give you a good starting point. From my experience the pointcloud from the photos should be more dense than the Lidar.
-
Here some screenshots:
The first is the point cloud obtained from Photoscan
-
Ponit cloud obtained via Lidar (ground + Buildings)
-
Lidar + Photogrammetry datum...
good correspondence between them..
-
Cross section along the center of the village...
You can notice a wrong z value for photogrammetric point cloud...
(It's the linear scale coloured one)
P.S.: Someone could tell me how post images in a single post?
thanks
-
Try turning the Lidar into a finescale DEM then use the x,y,z coordinates of the DEM surface as control/markers for the Photos. Its possible that either the Lidar or the camera IMU data is out. Try for a feature on the ground with a flat surface like path intersections depending on the Lidar density you should get it to within 10cm or so in the x,y and very accurate in the z.
-
Maybe this is the problem...
After align photos, appear NC
Does miss any EXIF data?
I don't understand...
-
Hi Piste
Could you provide which camera/lens did you use?
And you need at least the focal from camera to put in the PS to work properly.
-
Camera/lens used during the flight are provided
in the screenshot posted yesterday...
As you can see from EXIF it is a Hasselblad CF-39
with 50 mm focal lenght
-
Hello piste,
Could you please provide the initial and refined calibration data from the Camera Calibration dialog (Tools menu)?
In case they differ greatly we recommend to input refined fx, fy, cx, cy values as initial ones for all the images and realign the image set.
Having camera coordinates is not required, so you can use Generic pre-selection to speed up the process for large amount of images.
Also we recommend to optimize the photo alignment using GCPs with known coordinates.
-
Hi Alexey.
Here to you parameters that you asked to me...
Do I have to replace them with the second ones?
-
Hi Alexey,
I redo the entire project.
This time I used less photos,
Align with Generic, high accuracy.
Build geometry with Medium quality, height field, Sharp, face count 0, filter 0, hole 0.1
Texture orthophoto, mosaic.
you can see the related point cloud below in rtrue colours.
The white one is Lidar point cloud.
This time I didn't use any GCP but only scaled the model with a reference distance.
i obtained a better result as you can see in the cetion below.
But it's not completely acceptable in term of precision.
-
Other screenshots from the previous project...
This time I created an Ultra High quality model.
First of all here is Lidar point cloud...
-
Now, photogrammtery point cloud exported from Photoscan..
-
Lidar point cloud overlying the photogrammetry one..
-
A cross section through point cloud...
In this part of the point cloud there is a difference between Lidar datum and photogrammetric one of about 0,20/0.30 meters...
Flat surfaces show this singular artefact...similar to some waves..
Why?
-
Model view
-
Camera calibration parameters...initial
-
camera calibration parameters...refined
-
Model in perspective view.
You can see two marker used to set reference distance.
This distance was measured from an available ortofoto in a GIS software..
-
A screenshot of error panel...
What's the meaning of those values?
-
Alexey if you send me your email I'll send to you original images used in project.
In so doing I hope to obtained better results :P
Many Thanks
Best regards