Understanding Structured Light Scanning Measurements

Structured light scanning is used to create a 3D rendering of a physical object. The structured light scanner contains a projector which projects a light pattern onto the test object and one or more cameras which capture the distorted pattern created by the object’s surface (two cameras are typical). The images captured by the cameras are analyzed to identify the object’s shape. Using the calibrated spatial relationship between the projector and cameras, a point cloud is generated showing the exterior surface of the object by stitching multiple images from different positions around the object. The point cloud is polygonized to generate a mesh, which can then be exported as a stereolithography (STL) file for analysis, reverse engineering, stock model definition for CNC machining paths, in-process monitoring and location selection for material repair by additive processes, to name a few examples.

To make the best use of part dimensions extracted from structured light scans, we must understand the sensor performance. One metric used to quantify this performance is measurement reproducibility, or the “closeness of the agreement between the results of measurements of the same measurand carried out under changed conditions of measurement,” according to NIST. In this study, we used the artifact shown in Figure 1 to assess reproducibility by selecting 15 positions to create a mesh and repeating the 15-position measurement sequence 10 times using different positions for each mesh construction. The mean, range and standard deviation were calculated from the distributions in the feature dimensions. This incorporated the effects of both scanning and the position and orientation of the part relative to the scanner. This sequence was repeated for 10 positions and five positions to evaluate the corresponding sensitivity.

The Zeiss ATOS Q structured light scanner used in this study collects eight million points per scan with two cameras and an LED-based structured light projector. Measurement areas range from 100 mm × 70 mm to 500 mm × 370 mm. The measuring area applied in this study was 350 mm × 260 mm and the working distance was 490 mm. The distances between points varied from 0.04 mm to 0.15 mm. The artifact was CNC machined from 50.8 mm × 50.8 mm × 101.6 mm 6061-T6 aluminum stock. As shown in Figure 1, the geometry included 10 holes with nominal diameters from 2.9 – 15.9 mm and four step heights of 20, 25, 30 and 35 mm. After machining, the prismatic part was grit blasted to a satin finish.

Figure 2 displays a comparison of the diameter standard deviations as a function of mean circle diameter. The standard deviation values decrease with increasing circle diameter. This is because the larger circles have more points available to define their shape from the scans. Also, reducing the number of scanning positions increases the standard deviation of the circle diameters.

The step height results are displayed in Figure 3. There is no clear trend in standard deviation with step height size. This is because the height was defined as the distance between two planes in each case and the plane size did not depend on the step height. An increase in standard deviation with a reduced number of scan positions is observed, however.

To summarize the structured light scanner measurement performance for the selected artifact and features, two plots are provided: one for the 10 circles (Figure 4) and one for the four step heights (Figure 5). The vertical axis in each plot gives the number of measurement positions and the horizontal axis gives the feature size. The height map shows the reproducibility (one standard deviation) from 10 separate tests for each of the feature size-number of position combinations. The units are micrometers.

The outcome is that for the circles, with a dependence on the number of points used to define the feature on its size, low standard deviations (5 µm or less) are only available for many measurement positions (10 to 15) and moderate to large circles (6 – 16 mm). For the step heights, on the other hand, there is no strong dependence on the feature size (i.e., the contours are approximately flat). This is because the artifact geometry provided approximately the same surface area for measurements regardless of the step height. Low standard deviations (5 µm or less) are available for many measurement positions (10 to 15).