Laser Sample Data Processing - JG Conversions, LLC

Laser Sample Data Processing

Most users of 3-D scan data need to be able to reduce the number of data points from the typical 500K – 10 Million down to a reasonable amount of data so downstream processing is not overwhelmed. There are a few ways to process this much data and extract the useful information. With all processing methods, the point data information or model we provide is aligned to real-world axes and has a useful origin.

1. We can give the client the complete mesh with every scan data point. If the mesh of data points covers the entire surface of the part (including all interior features) it is called “watertight.” The file format is called STL (Stereo Lithograph) which is commonly used by companies who make Rapid prototype parts. Here is a diagram of a scan of a part. Note that typical scans are at least 1000 dots per square inch and can go as dense as only .005 inch between each data sample.

2. We can scan the entire model at high resolution and use only every, say tenth, data point inside CAD. You might think this is a silly way to do things because you can scan the model in a lower resolution. This however, is just how some 3-D CAD programs process large input data clouds. This is called data point sub-sampling. You can see from the diagram that you could miss important part features. You can also see that there are extra unneeded data points in the flat areas. We do not do it this way or provide our clients with this sort of data except when specifically requested. In this drawing we just show taking every other sample for simplicity.

3. Another full automated way to reduce the amount of data is to do data point simplification. For this method, the scanner software (which also has local data surface direction data as well as the surrounding points’ xyz coordinates) is able to eliminate a data point if it is close to the same height as the surrounding data points. The software will remove the points in a flat area while leaving all points in corners of features or where the surface is rough. The operator manually adjusts how close the height needs to be so the desired number of data points is preserved. Rough surfaces means that there are not as many points left for real features. It is not be possible to completely capture all areas of the part’s features using just this method if the part is complex because the number of input points is limited by many 3-D CAD programs. This number of sample points does, however, provide enough places to validate all the CAD’s solid body features as they are built. We put this set of points into its own IGES layer (think of it as a sub-file,) so it can be processed in the 3-D CAD program separately. We give this data to the client as a modeling aid so they can construct their solid model in their 3-D CAD program

4. A manual way to extract useful data from the data points is to visually identify the features and generate profiles or cross-sections of those features. We capture all the points in the non-simplified scan data those points which intercept a plane that crosses that feature. We vary the thickness of the plane until there is enough data points to reconstruct the profile when viewed normal to the plane. Profiles are constructed for all interesting features including rotation cross-sections, extrude and loft cross sections. A human operator skilled in 3-D CAD modeling determines what is useful.

If there are more than 20K points of profiles or if there are 2 non-related profiles physically together that could get confused, they are separated into individual IGES layers so they can be imported into a different 3-D sketch in the CAD program. Each layer can be assigned a different color and the display of each layer can be turned off entirely so as to not clutter up the display. Note that this method does not actually generate the lines of the profiles, but just the points where the line should pass thru if there is no local surface distortion; say from part casting shrinkage or wear. Here are some projects showing the steps of the process: Thermo Cap, Throttle Body.

A client’s 3-D CAD operator or ABJ Conversions determines the design intent and draws the lines of the profile appropriately. The number of data points given to our clients can be imported into any version of SolidWorks.

As the part becomes too complex with many features needing profiles, it is necessary to model the part directly from the scan data because too many IGES layers are needed. Another layer of software is needed to interpret all the information as described below.

5. We actually process the data a different way when the client requests a complete watertight mesh but some areas of the part cannot be seen by our scanner. We use a program called RapidForm XOR from Inus (also called RapidWorks.) This program is designed just for the task we are doing, reverse engineering solid models directly from 3-D scan data. There is a way to mostly auto generate solid features that closely match the all scan data points using statistical methods, but the results are fair to poor for matching the design intent of the original part. Angles will be almost right, but just a bit off because, in real life actual parts are just a bit off. Dimensions will not be even numbers because the software has no information about the original design of the part. This does not actually matter for some non-machine tool application like scale modeling where it just needs to look right. All other applications will result in extreme user frustration and it is generally a horrible experience when some person actually goes in to edit the resulting SolidWorks model. The software works amazing well compensating for areas missing in the 3-D scan. We then generate a watertight mesh (in STL format) from the solid model of the part. The mesh is cleaner than the mesh obtained directly from the scanner. This is typically used only for rapid-prototype production of the part. Note that the angles are not quite true.

6. We also use the RapidForm XOR software if the client request a completed SolidWorks model of the part and we can scan at least some of the part. The better way to use this software for this situation is to set up the directions for each feature to match oriented axes, generate the feature, and then later edit what it generates to more closely match design intent. Next, we export the RapidForm model into SolidWorks and do further cleanup before sending the completed model to the client.