|
|
Home : EvalViewer : Overview :
 | Previous: An EvalViewer Workflow |
The Quick Render Cloud and Cloud Sections - Multiple Clouds functions are fast and view-independent, but don't work on all clouds due to Point Order Dependencies in the cloud data. Even when these functions work, the quality of the display and the sections may not always have as high quality as you might want.
View Meshes have no Point Order Dependencies whatsoever, but are somewhat view-dependent. They can be smoothed in two directions, whereas Smooth Cloud only smooths in one direction. They support reasonably quick surface curvature computations, boundary curve extraction, and fast quadtree-based polygon reduction. Access them from the Mesh button tab.
View meshes can be best described using an example. In the picture below are the results of a Cloud Quick Render on a head that is difficult to scan.
To create a view mesh, start with a set of cloud data.
Imagine you have a piece of door screen or cheesecloth or gauze. The wires or fibers form a dense mesh with a characteristic distance or spacing between them. This distance is called the View Mesh Sample Size. Imagine also that the cloth or screen can stretch over any object, but if it stretches too far - larger than the S(ide)-Filter Size - it will tear.
Next, imagine that the cloud points have some size so that you cannot push the cloth or screen through the points. The process of creating the view mesh is then very similar to draping or pushing the cloth/screen onto the cloud.
Unfortunately, you can only create view meshes from one direction at a time. In the following example, we use front, bottom, right, left, and top views to create five view meshes. This arrangement of views is not the best for this particular data set because of the surface shapes and orientations, but it works well as a general recipe for getting started.
1 Divide the data in half along its center plane. You can then mesh each half separately.
2 Click the Make button. The meshing process creates a shaded surface.
3 Click the Smooth button. This suppresses the roughness in the data from the scanner.
4 Click the Reduce button to simplify the mesh in flat areas.
5 Click the Compress button to finalize the process and modify the View Mesh Object permanently. It is now just a conventional polygonal object, with as few polygons as possible.
The figure below shows the results of creating, smoothing, reducing, compressing, and manually cropping the mesh created from a side view:
Carrying out the same procedure for the other four views (there is no surface data for the back of the head), we obtain the following model:
Notice the difference in the view-mesh front view compared to the quick-render front view.
As the previous pictures show, view meshes simply fill in areas where no data exists. You can control this filling with the Maximum Fill Size parameter, which is specified in units of samples.
The number of polygons corresponding to the two view mesh images is under 300,000 quadrilaterals, even though the original data set contained 1.3 million points. There are still large overlapping areas that can be eliminated manually with further effort.
Polygon reduction can occur in various ways. In this case, since view meshes are regularly spaced quadrilateral meshes, a "quadtree" simplification process is used. Each view mesh is cohesive and "water-tight" except for the visible holes that you can detect using Mesh > Boundary. (You can only use Boundary prior to Compress.) The picture below shows a close-up of the view mesh on the nose region of the head.
You can export compressed View Meshes as Wavefront OBJ, STL (stereo-lithography), Inventor, or DXF files, or an Alias wire file by using File > Save As > Save Polygons As. We recommend that you keep your polygon count below 100,000 whenever you write an Alias wire file.
After you create view meshes, section them using the surface sectioning operators. Do not use the cloud sectioning tools. The amount of overlap in the sections corresponds to the amount of overlap in the view meshes. If you use sections directly for building surfaces, you should find that, even though you spend more time creating the polygonal model, you spend less time "cleaning up" sections - that is, you won't need to join unconnected sections, delete bad points and bad sections, or apply other operators on the sections tab.
 | Next: EvalViewer cloud Workflow example 1 |