Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Topics - RalfH

Pages: [1] 2 3
General / orthophoto from single aerial image and DSM
« on: May 30, 2018, 03:58:50 PM »
Is it possible in Photoscan to create an orthophoto from as single aerial image, a DSM and some ground control points? Theoretically, it should work (camera position and orientation computable from ground control points, then projecting the image over the DSM) - but how?

General / Calculation of motion blur, ground resolution and image overlap
« on: January 10, 2014, 03:04:08 PM »
Hi all,

this open source tool may be of interest to some of you:

UAVphoto is a simple tool that allows to calculate parameters like ground pixel size, motion blur and sequential image overlap from flight parameters (velocity and altitude) and camera parameters (focal length, shutter time, image interval etc.).

General / Rolling shutter problem
« on: May 29, 2013, 11:14:14 AM »
As the rolling shutter problem has popped up in the discussions once in a while, here are some exaples of what it can look like:

When creating a dense point cloud, Photoscan appears to first load all images, leaves them untouched (for ultra) or resamples to 1/2 (for high), to 1/4 (for medium) etc. I assume that the resampled images are then undistorted (?). Judging from memory load, it looks as if the images are then all stored uncompressed in the RAM. This seems to be the main reason for the high memory load at high and ultra quality settings during point cloud creation.

Would it be possible to change memory use in a way that always only those photos are loaded into the RAM which are necessary for creating the current depth map? Of course this would slow down things when memory is not a limiting factor, so perhaps this functionality could be activated with a tick box. That way, creating high and ultra quality dense point clouds for large projects would be possible even with limited RAM (16 GB in my case).

Feature Requests / Turn off depth filtering
« on: May 03, 2013, 06:24:04 PM »
When creating a dense point cloud, the strength of the depth filtering can be set in the preferences to either "mild", "moderate" or "aggressive". "Mild" removes the lowest number of points which - as I understand it - do not appear to belong to a connected surface. When modelling a sparse plant canopy, "mild" results in much better results than "aggressive". Still, the upper parts of some plants are removed by depth filtering. The same problem will likely affect other  objects with complicated geometry.

It would be great if the next version of Photoscan would have a fourth option: to completely turn off depth filtering.

General / Turn off depth filtering?
« on: April 29, 2013, 11:53:53 PM »
When creating a dense point cloud, the strength of the depth filtering can be set in the preferences to either "mild", "moderate" or "aggressive". "Mild" removes the lowest number of points which - as I understand it - do not appear to belong to a connected surface. When modelling a sparse plant canopy, "mild" results in much better results than "aggressive". Still, the upper parts of some plants are removed by depth filtering.

Is there a way to completely turn off depth filtering?

At the moment, I am working a lot with dense point clouds. I'm still amazed by how - from nothing but a set of photographs - Photoscan can create and export dense point clouds with more than a 1000 million points without problems (yes, some patience is required). One thing I do with the dense 3D point clouds is, for example, to create 2D point density maps.

A feature that I think would be very useful would be an option to export a point reconstruction parameter as an additional column in the dense point cloud xyz file. That way I could also create maps showing the spatial distribution of point reconstruction quality.

P.S.: Of course, this feature would also allow to do point cloud filtering by reconstruction quality in post-processing.

General / When does depth filtering happen?
« on: March 19, 2013, 12:28:49 PM »
At what point in the 3D processing do the depth filtering settings (mild, moderate, agressive) play a role?

The reson why I ask this is the following: I have a project with approx. 450 photos of very sparse plant cover over a ground surface. Because the plant cover is so sparse, Photoscan finds many points on the ground, so I can theoretically use the dense Photoscan point cloud like a lidar point cloud to measure vegetation height. I have now computed depth maps and dense point clouds at different quality settings (lowest to high) but always at agressive depth filtering setting. When looking at the dense point clouds I have the impression that very many plant 3D points are missing, and I assume that this is because of the "agressive" depth filtering.

If I want to compute dense point clouds at "mild" depth filtering, do I have to re-compute the depth maps? It all comes down to when the depth filtering is applied by Photoscan: during depth map computation or during dense point cloud computation from the depth maps?

I recently created a project with two chunks and a few dozen photos per chunk. After importing photos (but before alignment) I thougth that maybe I should try to process all images as one chunk, so I just moved all photos from the second chunk to the first, using drag and drop. Suddenly, the photos which had been moved did not have camera information anymore ("NC" after the file name), and when I tried to align photos, no alignment of any photos was done. I had to remove the "NC" photos and re-import them (then they did not have "NC" anymore), and afterwards all images were aligned without problems. I think something goes wrong during drad and drop.

Feature Requests / rapid masking of bluescreen background
« on: March 08, 2013, 05:38:59 PM »
I am at the moment experimenting with using the bluescreen technique to facilitate masking when photographing small objects: I have a background which is blue (actually a fullscreen blue image displayed on a computer monitor to ensure even colour and brightness and to avoid any shadowing on the background), and my objects don't contain blue colours. This makes masking with the magic wand quite easy.

For large numbers of photographs (in my case, 100-150 per object), there is still a lot of manual work involved in masking the blue background in all photographs. I think it would be great if there was a button to "mask blue in all images", again with a tolerance slider as in the magic wand tool.

General / No depth maps for some camera positions
« on: March 01, 2013, 03:11:56 PM »
I have a problem with a project for which I walked around an object to capture it from all sides. Doing this had the result that some camera positions are "seen" by other cameras. As Alexey told me a while ago, depth maps for these cameras are not computed, and these cameras are not used for dense point cloud or meshed model reconstruction. This is very unfortunate, because I now have a set of photographs showing the object from all sides, but I have an area in my model for which I have no dense point cloud coverage. Does anybody know a way (or a workaround) to force PhotoScan to compute depth maps for all camera positions?

Bug Reports / Point-based chunk alignment
« on: February 28, 2013, 08:37:58 PM »
In version 0.9.1, when I use point-based chunk alignment, PhotoScan first detects matching points even though they already exist from photo alignment. I think this is a waste of processing time.

General / Rotating camera poses?
« on: February 05, 2013, 11:25:36 AM »
I would like to rotate the camera poses by exporting the cameras.xml file, edit the values in the transformations section and re-import the file. Rotating the XYZ camera coordinates is not a problem (multiply with a rotation matrix), but does anybody know how to rotate the r11, r12, r13... parameters by the same angles?

Feature Requests / Thousands and millions separator, gradual selection
« on: February 04, 2013, 05:42:54 PM »
Three minor requests:

For the numbers of points and faces in general and for the numbers of points and selected points displayed when working with gradual selection in particular, it would be great to have a thousands and millions separator (,) to make reading of long numbers easier.

For the gradual selection it would be great if the percentage of selected points could be displayed (e.g. "points: 4,234,123 selection: 211,623 (5 %)") to make it easier to judge how strongly the point cloud is affected.

Also, in the gradual selection dialogue it would be great if I could specify something like: "select the 20% of points with the highest reprojection error" or "select the 5 % of points with the highest reconstruction uncertainty".

Bug Reports / Cancel during "generating mesh"
« on: February 04, 2013, 05:13:12 PM »
When trying to cancel the "generating mesh" processing in batch processing mode, PhotoScan 0.9.1 (and I believe also earlier versions) asks for confirmation but then keeps on processing. Pressing Escape or clicking on the "close" button in the upper right corner of the batch processing window has the same result - I can't stop the processing. When pressing "Pause", the button becomes inactive for half a second and then becomes active again; the processing doesn't pause.

Pages: [1] 2 3