deleting Tie points

Dennis_Nikolaev · June 21, 2015, 2:35am

once in a while Pix4d places a tie point in an image where it is not visible. I cant find a way to tell the software that this image should not have this tie point. Same thing with the manual tie point. If I click it by accident - no way to remove it completely from the photo.

Dennis_Nikolaev · June 21, 2015, 3:48am

another question: Is multiscale processing help to avoid gross mistakes but adds noise?

nikoleta.guetcheva · July 1, 2015, 3:34pm

Dear Dennis,

When an automatic tie point or a manual tie point is not correctly reprojected on all the images that it appears, then this means that the reconstruction of the model is not good and needs further investigation. This may be caused due to many reasons like low overlap, bad camera model, bad quality images etc.

If you mark a manual tie point on an image that you did not want to, you can delete the mark by posing the mouse over this image and clicking Delete.

Best regards,

Dennis_Nikolaev · July 2, 2015, 1:43pm

that’s helpful.
Thank you

Swampy · September 1, 2015, 8:07am

Related to this topic:
I process mainly imagery of vegetation only, often with sparse canopies or tall grasses. Often the X,Y values of my Manual Tie Points lie on the ground under the canopy of trees for many of the images. I therefore do not select points (make a yellow circle) in those images, but I am concerned that the height of the canopy (Z value of the green circle location) may be considered in calculating the optimized elevation. I would prefer to be able to tell the software not to use that image at all just for that specific MTP optimization. Is this a valid concern, and if so, would this be a possible option in future?

Felix_Archambault · September 29, 2015, 2:06pm

What I do is the following: I make sure to identify the correct place on a few pictures. Then, I click apply and project on all images. I then only click on images where the projected point is not where it should be. If I cannot see where the point should be, I just ignore the image. I think that would just confuse the software. So I just leave those be. Works well so far

Tamas_Csontos · August 31, 2016, 1:03pm

Hi,

Is there any progress on this?

It would be a really-really usefull additions to the capabilities.

I had several projects where deleting a few ATP per picture would solve the problems.

Thanks,

Tom.

Pierangelo_Rothenbuhler · March 31, 2017, 10:41am

Hi Swampy, Felix and Tamas,

@Swampy: We are not sure if you mean Manual Tie Points (MTPs) or Automatic Tie Points (ATPs).
MTPs are points that are manually added to an existing point cloud by marking a specific feature in images. For example, you could mark an MTP on a feature that is on the ground below a tree. The feature should be marked on images where the feature is visible. If there are images where the feature is visible, but the green cross is not at the correct place, the feature should be marked at the correct place in the image. This will give additional information to the software to correctly create the point cloud, e.g. adjust the Z value of the feature and surroundings. If there are images that are associated to the MTP in which the feature is not visible, no additional mark should be created as Felix has described in his workflow and these images will not be taken into account.
More information on how MTPs can be marked in this article: https://support.pix4d.com/hc/en-us/articles/202560349

On the other hand, ATPs are the result of step 1. Initial Processing. All ATPs together are a point cloud which is the basis for all further computations in step 2 and 3 of processing. These can be influenced by MTPs, Processing Options, camera model, flight plan and many more parameters.

@Tamas: There are many filtering steps during calibration that should ensure that there are no wrong associations between an image and an ATP. If we correctly understood, you would like to have the possibility to remove certain ATPs if they appear at a wrong place. Please correct us if this is not what you meant.

For the moment the automatic tie points cannot be edited. Instead, parameters such as the flight plan (overlap, height,…) and Processing Options should be modified to improve the reconstruction. If certain ATPs are wrong, it could point to other issues in the project, which should be addressed on a project level by editing the parameters mentioned above.

Regards

Chad_Masterson · May 2, 2017, 1:41pm

I use this software for surveying quarries and measuring stockpiles and the ability to correct Automatic Tie Points or simply delete them would be very useful. On more than one project, I have noticed that the software places the ATP’s in images where the location is not visible (I like to use paint marks around the base of the piles to add in Manual Tie Points). It is understandable that when you are taking pictures of a large stockpile of rock that it may be difficult for the software to always locate the correct position of a tie point. However, the software’s inability to locate the correct position of an ATP, coupled with the fact that ATP’s cannot be edited, creates a lot of problems that could be more easily corrected. Do you have any intentions of adding ATP editing? From what I have noticed, adding MTP’s does help, but the original ATP’s still bring a lot of error to the project.

Thanks,

Chad

Pierangelo_Rothenbuhler · May 4, 2017, 9:53am

Chad,

For the moment there are no plans to include ATP editing. It should be possible to correct such problems by adapting the flight plan, the processing options and adding Manual Tie Points (MTPs).

For measuring stockpiles, this could mean to use a double grid flight plan with oblique cameras, such as described here: https://support.pix4d.com/hc/en-us/articles/202557459#label8

As you mentioned, stockpiles can be difficult to reconstruct (e.g. coal, rocks, sawdust…), because the content looks very similar. In such cases, it can happen that the surface of the stockpile is “doubled”, because the software cannot match certain images between each other. For these cases, we would recommend to add several Manual Tie Points (MTPs) in the problematic areas and to Process > Reoptimize the project. If necessary, you could also try Process > Rematch and Optimize, which would change the ATPs that were created in the first run. Note that this can take more processing time than the reoptimize procedure.

Hope this helps!

Eric_Braun · June 30, 2017, 1:04am

In my case some of the cameras have turned upside down and created tie-points above the land surface and then proceed to densified that surface, making a big blob in the sky. The area that caused this is a confusing tree canopy that we’re not really interested in. It would be great if we could just remove the tie-points in the sky. I know I can remove the upward facing cameras, but the tie points remain. Would removal of these cameras keep the big blob in the sky from being created later in the process?

Pierangelo_Rothenbuhler · July 3, 2017, 2:11pm

Eric,

This probably means that the calibration of the cameras during step 1. Initial Processing did not go well. We would check the quality report of the project, e.g. that the camera optimization is below 5%. Adding a few Manual Tie Points (MTPs) and Process > Reoptimize could help to orient the cameras correctly as well.

If the quality report looks okay (green crosses in the summary) and the reconstruction looks good in the rayCloud view, then you could delete points from the densified point cloud using the point cloud editing tool: https://support.pix4d.com/hc/en-us/articles/202560499. Please first make sure that step 1. Initial Processing went well and correctly calibrated most of the cameras.