# Apply terrain filter to classified point clouds

**URL:** <https://community.pix4d.com/t/apply-terrain-filter-to-classified-point-clouds/24091>\
**Category:** PIX4Dsurvey Feature Request\
**Created:** [September 29, 2022, 4:13am UTC](https://community.pix4d.com/t/apply-terrain-filter-to-classified-point-clouds/24091 "2022-09-29T04:13:34Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![PPGroup](https://avatars.discourse-cdn.com/v4/letter/p/7ea924/32.png) [@PPGroup](https://community.pix4d.com/u/PPGroup)\
**Post date:** [September 29, 2022, 4:13am UTC](https://community.pix4d.com/t/apply-terrain-filter-to-classified-point-clouds/24091/1 "2022-09-29T04:13:34Z")

</div>

Would it be possible to run the terrain filter based off the point cloud classifications? Pix4d mapper does a good job of automatically classifying the point cloud, although it does miss some stuff, especially with building roofs and sometimes vehicles. These missed points remain as ground classified and can take a lot of work to find and reclassify manually. Would it be possible to add a filter to the Terrain filter so you can just run it on selected classes of point cloud. From tests I’ve done with a workaround this removes almost all the incorrectly classified ground points very efficiently

---

<div class="post-metadata">

**Author:** ![Daniele\_Lecci](https://sea2.discourse-cdn.com/pix4d/user_avatar/community.pix4d.com/daniele_lecci/32/9493_2.png) [@Daniele\_Lecci](https://community.pix4d.com/u/Daniele_Lecci)\
**Post date:** [September 29, 2022, 1:20pm UTC](https://community.pix4d.com/t/apply-terrain-filter-to-classified-point-clouds/24091/2 "2022-09-29T13:20:08Z")

</div>

Hi @PPGroup,

Thank you for your message. Did you already try to change the _terrain filer_ settings to see if you can achieve your goal with fewer steps?

Otherwise, once you have run the terrain filter the first time, you can export the specific class where you want to run the filter again.

![image](https://global.discourse-cdn.com/pix4d/original/2X/f/f5fa7f0e59f7595f3e581a5d00b0e4c168097266.png)

Once exported, you can import this point cloud to PIX4Dsurvey and run the terrain filter on this point cloud only by selecting it as _Input Point Cloud_ in the _Terrain Filter_.

![image](https://global.discourse-cdn.com/pix4d/original/2X/7/771a3d3e0df6d5439ed4a36b45d22eb149a1326b.png)

I also have notified the PIX4Dsurvey product team for eventual consideration 😉.

Let me know if this helps you.

---

<div class="post-metadata">

**Author:** ![PPGroup](https://avatars.discourse-cdn.com/v4/letter/p/7ea924/32.png) [@PPGroup](https://community.pix4d.com/u/PPGroup)\
**Post date:** [October 18, 2022, 9:21pm UTC](https://community.pix4d.com/t/apply-terrain-filter-to-classified-point-clouds/24091/3 "2022-10-18T21:21:09Z")

</div>

Yeah i have, the best results i get are from exporting a point cloud from mapper with just the ground classified points and then running terrain filter in Survey with a low above ground threshold to get rid of any small bushes, lumps of grass, bottoms of pole/fences/objects.

This workflow however is not possible when using Matic & your suggestion of using the terrain filter multiple times would potentially do a similar job

---

<div class="post-metadata">

**Author:** ![Daniele\_Lecci](https://sea2.discourse-cdn.com/pix4d/user_avatar/community.pix4d.com/daniele_lecci/32/9493_2.png) [@Daniele\_Lecci](https://community.pix4d.com/u/Daniele_Lecci)\
**Post date:** [October 19, 2022, 12:28pm UTC](https://community.pix4d.com/t/apply-terrain-filter-to-classified-point-clouds/24091/4 "2022-10-19T12:28:10Z")

</div>

Hi @PPGroup,

You are right; using PIX4Dmatic it is impossible to follow the same procedure because PIX4Dmatic does not classify the point cloud.

Using the terrain filter in PIX4Dsurvey the first time would help you simulate the ground classified point cloud you are getting with PIX4Dmapper; running a second time the terrain filter in PIX4Dsurvey (this time fine tuned) only on the ground classified points would potentially do a similar job, as you stated in your previous post.