Support Website Contact Support Blog

Initial Processing Automatic Tie Point filtering


Does any one know if there is any way to filter or manually delete Automatic Tie Points that are outliers?
If that is not possible I would like to suggest the inclusion of a method to manually delete points (much like it can be done in the dense cloud) as well as a way to select points with large standard deviations.



Filtering points would be really helpful in some cases. I totally agree with you. Does anyone know how to do it?

The reason I was pointing this out is because Pix4D seems to have real troubles handling trees compared to other phogrammetric software, I have seen flights where the external orientation is completely messed up because most of the picture (but not all) falls on large green trees (this did not happened on other softwares with the same pictures).
I believe that the ability to filter, both manually (mainly outliers) and automatically (standard deviations) would be handy.



Hi all,

Currently, you can not edit directly the automatic tie points. You can improve the Initial Processing (calibration), by adding some manual tie points: but you can not delete the automatic tie points.

It is actually a very good idea. I have added it in our Suggestion list so that our Product Development team will consider it for future version of the software.

@Isaac H. E.: Mapping trees areas is not easy, this is why we recommend a special flight plan and specific processing options for such areas. For more information:

Best regards,

It’s been 4 months, Has anyone figured out a workaround yet ?

I’m on Version 2.0.81 and having significant problems with not being able to remove the Automatic Tie Points to generate a true DTM.  

My Workflow is to use “Point Cloud Classification” In Step 2 Advanced, then  “Generate DTM (beta)”, then clean up the dense Point Cloud by assigning any objects or vegetation missed by DTM routine to the “Deleted” Point Group, and re-run Step 3 as per instructions.    

The Automatic Tie Points are still being used to create contours.  This defeats all the manual work done to correct the DTM.    

Can you please add the Option in Step 3 (DSM Filters) to not use Automatic Tie Points, or add a workflow for DTM?

Is the only current solution to draw surfaces over ALL Automatic Tie Points created on Objects and Vegetation & Re-Optimize ?

I must be missing something important.  

Thanks in advance for any help,


I agree completely.

Pix4d says to move all classified points into the deleted group and then regenerate DSM. We are hoping that Pix4D

will integrate and option to choose point groups to be included or excluded in the DSM generation. Hopefully sooner than later.


The Automatic Tie Points are not taken into account to generate the DSM or the DTM, only the points coming from the densification process. As such, even if you cannot classify them, they will not degrade the results you will get.

However, Automatic Tie Points that seem wrong should be taken as a warning flag: they typically indicate that something went awry. It is a good idea to add Manual Tie Points as per Christina’s suggestion: this improves the point cloud. 

Choosing which groups are included for generating the outputs of the 3rd step is an interesting suggestion, as well as identifying points that are outliers. We will add them to the list of suggested features.

Best regards,

Mismatched tie points tend to happen quite often. This is the case for every photogrammetric software I’ve used. Most other software I’ve used, however, has the option to do an automatic blunder detection and removal, and to manually remove bad tie points and run a bundle adjustment on what remains. While I understand that adding manual tie points may improve the overall quality, this is simply not practical for large AOIs with hundreds or thousands of photos.

I believe implementing blunder detection and tie point removal would go a long way to improving the software’s capabilities.

Thanks, looking forward to this implementation.


For now,I cant manually delete Automatic Tie Points that are outliers too,the version is Pix4Ddiscovery 3.0.13.

Hi Jim,

It is not possible to delete automatic tie points (ATPs).

In order to improve the calibration and so the location of ATPs, you could add manual tie points (MTPs) to better align the images and so the points in the rayCloud. The MTPs should be added in the problematic areas where there is for instance little overlap and few connections in the model or where you see misalignment. To know where to add a MTP, you should verify the re-projection of a point. Click different locations in the point model and check on the images (thumbnails in the right sidebar) if the point is re-projected at the correct location on all images.

How to add and mark MTPs in the rayCloud:

After adding the MTPs it will be needed to rematch and optimize the calibration of step 1. This means all subsequent results (steps 2 and 3) will be removed. For more information:

After processing step 1, it is possible to draw and apply a processing area (3D polygon) that will discard points lying outside. You can adjust the horizontal boundaries but also the upper and lower limits. For more information:



I’ve got the same problem here. I have an outlier tie point due to a slow moving vehicle :slight_smile:

I guess no manual tie point will fix it. 

Normally projections errors are bad calibrations results but sometimes they are only noise. It would be nice to get rid of them.



Dear Diego,

Yes, you are right that it is not very pleasing to have this outlier point. However, since it is just a single point, it will have no influence on further processing. 

Best regards,

Hi everyone,

Just wondering if there has been update on the ability to mask/remove inaccurate ATPs. As you can see from this image, Initial processing is still generating a number of ATPs that are simply not possible.

I have tried adjusting processing options including Geometrically Verified Matching as well as Accurate Geo-location and Orientation but cannot find a solution to this.

Does this impact the quality of the mesh or the densified cloud point? 

Hi Rick,

If the sparse point cloud is noisy (after step 1), it is recommended to add MTPs as suggested in the comment above (Nov. 4th). It is the best way to correct for the geolocation of points.

If these undesired points are too numerous, they will impact the densified point cloud and mesh generated during step 2.

To improve the results, you can also check the processing options of step 1:
In your case we would suggest to apply the template 3D Models given the type of project:


Good morning,

I am monitoring a landslide body at different times.

Since the landslide is a dynamic phenomenon, I would like to eliminate the ATPs located in the central area (I do not want moving points defined as ATPs), or to tell the software not to look for ATPs in that area

Is it possible?


Hey Filiberto, 

The Automatic Tie Points (ATPs) are generated from the keypoints that are automatically detected in the images, thus it will not be possible to remove them from the model. You could decrease the number of keypoints to be extracted per image following this article, however, it will not be possible to eliminate the keypoints (and eventually the ATPs) at a specific area.