# ICP is not working for specific data

**URL:** <https://discourse.vtk.org/t/icp-is-not-working-for-specific-data/3670>\
**Category:** Development\
**Created:** [June 29, 2020, 11:06pm UTC](https://discourse.vtk.org/t/icp-is-not-working-for-specific-data/3670 "2020-06-29T23:06:11Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![siaeleni](https://discourse.vtk.org/letter_avatar_proxy/v4/letter/s/b4bc9f/32.png) [@siaeleni](https://discourse.vtk.org/u/siaeleni)\
**Post date:** [June 29, 2020, 11:06pm UTC](https://discourse.vtk.org/t/icp-is-not-working-for-specific-data/3670/1 "2020-06-29T23:06:11Z")

</div>

Hi all,  
I an using ICP (Iterative Closest Point algorithm - vtkIterativeClosestPointTransform) and so far works great for stl data that I have.  
But I just received some new stl files and it doesn’t give correct results, even the surface looks very good.

 ![Capture](https://discourse.vtk.org/uploads/default/original/2X/3/304308493d7b93ec23bf75b4f1aa87d7bebf35a5.jpeg)  
Do you think that bad triangulation would cause these wrong results? I tried to re-generate my stl with better triangulation, but still, ICP is not working as expected.

Do you have any idea that can help me investigate further?

---

<div class="post-metadata">

**Author:** ![pieper](https://discourse.vtk.org/user_avatar/discourse.vtk.org/pieper/32/17_2.png) [@pieper](https://discourse.vtk.org/u/pieper)\
**Post date:** [June 30, 2020, 7:43pm UTC](https://discourse.vtk.org/t/icp-is-not-working-for-specific-data/3670/2 "2020-06-30T19:43:11Z")

</div>

Hi @siaeleni - probably it also depends a lot on initialization - are the new datasets close before you run the ICP? Maybe you can share examples of data that do and don’t register well?

I haven’t used ICP a lot myself, but I plan to soon so I’d also be interested in hearing any advice about what works well.

---

<div class="post-metadata">

**Author:** ![siaeleni](https://discourse.vtk.org/letter_avatar_proxy/v4/letter/s/b4bc9f/32.png) [@siaeleni](https://discourse.vtk.org/u/siaeleni)\
**Post date:** [July 6, 2020, 6:48pm UTC](https://discourse.vtk.org/t/icp-is-not-working-for-specific-data/3670/3 "2020-07-06T18:48:24Z")

</div>

Hi @pieper,  
Exactly, you are right. My data are generated from different people with different software,  
thus these after the initial guess (after the translation) are not very close.

Probably a pre-process step is needed here.