Handling Point Clouds with Grasshopper/Rhino

What is the best way to handle point clouds with Grasshopper/Rhino.

I wish to achieve the following in the most computer resource friendly way possible;

-Handling ¬ 2,000,000+ points.

-Coloring individually using a sample picture.

-Baking the colored point cloud.

The X,Y,Z data of the points is being read by reading a text file with the information in csv format.

Each point is to be colored individually using a sample picture.

With this simple setup I am showing I managed 800,000 points with 9Gb of my computers Memory being utilized.

To achieve this sort of 3D survey output;

3D surveys usually contain millions of points and specialized software only can handle that volume of points. But I am looking into ways of manipulating all this data directly inside grasshopper. Ideas/suggestions are more then welcome.

- How can you bake these points with the color information stored with the point?

Each point would have its color set by object - and it retains the color of the custom previewer. (What you are seeing in the picture is the grasshopper preview)

- Are there settings that I can tweak for Rhino/Grasshopper to handle this volume of points more efficiently?

Replies to This Discussion

Permalink Reply by David Rutten on December 11, 2014 at 4:00pm

Grasshopper doesn't support pointclouds as a native data type. Each point must be stored separately, which greatly increases the memory footprint. It's also not possible to associate a colour with each point, except if you're willing to keep a list of synchronized colours in parallel with the point list.

For this sort of geometry Grasshopper is almost certainly not your application of choice. What you need is something that knows how to squeeze every last inch of performance out of huge point sets.

Apart from displaying, what other things do you want to do with this data?

Permalink Reply by David Rutten on December 11, 2014 at 4:13pm

Incidentally, a single point entry should require 3 * 8 bytes (assuming points are three-dimensional and stored as double precision floating point numbers) plus 3*1 bytes (assuming colour is stored as byte-accuracy red, green, blue channels) = 27 bytes. More likely 28 bytes as colours are probably stored as 32-bit integers, allowing for an unused alpha channel.

28 * 800,000 equals roughly 22 megabytes, which is way down from 9 gigabytes. That's a 400 fold memory overhead, which is pretty hefty.

Grasshopper stores points as instances of classes, so on 64-bit systems it actually takes 64+64+3*8 = 152 bytes per point*, which adds up to 122MB, still way less than 9GB. It would be interesting to know where all the memory goes...

* Grasshopper points also store reference data, in case they come from the Rhino document. This data will not exist, but even so it will require 64-bits of storage.

Permalink Reply by Kane Borg on December 14, 2014 at 2:33pm

Hi David,

Thanks for your input on this,

I would like to manipulate the point cloud to be able to select ranges x,y,z data, to effectively cut slices through the model and get sections out of them. In general it would be very useful to be able to manipulate real 3D surveys to use as context for sculpting apertures in buildings, visualizing in-city views accurately etc..

The x,y information are real world coordinates so a gps point of a person inside this particular space would appear referenced as a dot. I wish to use the survey and the proximity to this dot to manipulate survey points in the area, such as creating visual representations of people flowing through the surveyed Area. This makes most sense on a macro space like a 1km^2. But if this is not possible I can always reduce the extents of the observable space.

What would be an effective way for me to diagnose where effectively Memory is being used up?

Permalink Reply by Andres Rivera on July 6, 2017 at 4:01am

I see this thread is a bit old, but maybe this plugin helps for future pointcloud queries within gh! Very easy and fast to use http://www.grasshopper3d.com/group/tarsier .

RSS