[pdal] PCLBlock drops non-XYZ dimensions & what to do about

[Michael Rosen]

this is very helpful. Thanks for passing along some of the backstory.  The
visualization comment caught my eye:

>writers.pclvisualizer

>Exists as a writer simply so that we can stick it on the end of a pipeline
and visualize the results. But you >probably have a tool to view the
results anyway, and this doesn't write anything to disk (should you be
happy with >your result and want to save it).

Actually, I don't have such a tool just yet.  Hobu uses plas.io but ...
uploading a file to a web app seems impractical for me.  What do you use?
I'm on a LinuxMint environment.


msr



On Wed, Mar 15, 2017 at 5:56 AM, Bradley Chambers <brad.chambers at gmail.com>
wrote:

> Michael,
>
> Also, if you take a peek at the PCL plugin prior to our refactor (proposed
> [1] and implemented [2]), you'll see some good examples of doing exactly
> what you propose wrt extracting indices in many of the filters, for example
> [3] and [4].
>
> Brad
>
> [1] https://github.com/PDAL/PDAL/issues/1404
> [2] https://github.com/PDAL/PDAL/pull/1432
> [3] https://github.com/PDAL/PDAL/blob/006ea71362a5486109f48d2981b72a
> da15b31916/plugins/pcl/filters/DartSampleFilter.cpp#L84-L98
> [4] https://github.com/PDAL/PDAL/blob/d32c77f37007a494ce88e1e9ca64a1
> 4a6b21892f/plugins/pcl/filters/GroundFilter.cpp#L99-L161
>
>
>
> On Tue, Mar 14, 2017 at 8:53 PM Bradley Chambers <brad.chambers at gmail.com>
> wrote:
>
>> Michael,
>>
>> You are correct that there is no out of the box LAS point type, although
>> you could in theory create one [1].
>>
>> Various PCL modules provide methods such as "getInliers" to get a vector
>> of indices [2]. PCL treatment of inliers varies somewhat from module to
>> module, so one has to read the docs carefully.
>>
>> For an example of acting on the indices, you can look at the
>> OutlierFilter where we start by processing a PointView to obtain (in this
>> case) two vectors of PointIds, the inliers and the outliers [3]. We later
>> create a new output PointView by appending only the inliers from the input
>> PointView [4].
>>
>> No need to sort. The PDAL to PCL conversion [5] is order preserving, so
>> unless your PCL code is doing something to the order, you should be good.
>>
>> Brad
>>
>> [1] http://pointclouds.org/documentation/tutorials/
>> adding_custom_ptype.php#how-to-add-a-new-pointt-type
>> [2] http://docs.pointclouds.org/1.7.1/classpcl_1_1_sample_consensus.html
>> [3] https://github.com/PDAL/PDAL/blob/master/filters/
>> OutlierFilter.cpp#L159
>> [4] https://github.com/PDAL/PDAL/blob/master/filters/
>> OutlierFilter.cpp#L203-L209
>> [5] https://github.com/PDAL/PDAL/blob/master/plugins/pcl/
>> PCLConversions.hpp#L127-L191
>>
>>
>> On Tue, Mar 14, 2017 at 6:27 PM Michael Rosen <michael.rosen at gmail.com>
>> wrote:
>>
>> Picking up on a conversation started on Gitter:
>> https://gitter.im/PDAL/PDAL
>>
>> One problem we have with the PCLBlock filter is that we currently drop
>> all non-XYZ dimensions because when we marshal point cloud data from PDAL
>> --> PCL we needs to specify a PCL-compatible point type and we pick XYZ.
>> As a result, when we marshal the result back, we there is nothing except
>> XYZ.
>>
>> One workaround suggested was to create a PCL point type that fits the
>> data.  My quick look suggests there is not such point type, even for the
>> simplest of LAS files.  Can someone confirm this?  These, I think, are the
>> options http://docs.pointclouds.org/1.5.1/point__types_8hpp_source.html
>> ? So, For example, no way to carry Classification to say nothing of LAS
>> metadata like the array of "Number of points by return."
>>
>> The other suggestion was to return just the indices of the desired
>> points.  Can someone say a little more about what PCL API would enable
>> this?  Once we had it, how can we get PDAL to remove all the points except
>> those returned?
>>
>> One thought I had would be to sort the point cloud before marshaling it
>> to PCL and similarly sort the result from PCL.  Create the desired result
>> set by marching through the point clouds in parallel removing anything from
>> the original point cloud that isn't in the PCL result.
>>
>>
>>
>> msr
>> _______________________________________________
>> pdal mailing list
>> pdal at lists.osgeo.org
>> https://lists.osgeo.org/mailman/listinfo/pdal
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/pdal/attachments/20170315/4ed72706/attachment.html>