Featured Product
This Week in Quality Digest Live
Innovation Features
Nicola Olivetti
Digital visual management puts visual tools into the hands of team members across multiple sites
Venkat Viswanathan
It’s not just about energy
Steven Brand
Building better, smarter, and more secure supply chains
Matthew M. Lowe
Confronting the fear of being displaced by machines
Tara García Mathewson
Cognitive strategies instruction moves students toward achieving knowledge transformation

More Features

Innovation News
Deep Reality Viewer creates stereo, high-definition 3D images without using a monitor
Wing’s shape shifts automatically for aerodynamic loads (e.g., takeoff, landing, cruising)
Process concerns technology feasibility, commercial potential, and transition to marketplace
Identifying the 252 needs for workforce development to meet our future is a complex, wicked, and urgent problem
How established companies turn the tables on digital disruptors
Building organizational capability and capacity to create outcomes that matter most
‘What if we could develop electronic systems that we wrap around a bridge or cover the walls of our office?’
System breaks down complex designs into easily modifiable shapes for custom manufacturing and 3D printing
How to solve any problem and make the best decisions by shifting creative mindsets

More News

Amir Grinboim


Acquiring Richer Point Cloud Data

What makes the BLAZE 600M 3D optical scanner faster and more accurate

Published: Tuesday, August 23, 2016 - 13:49

Hexagon Manufacturing Intelligence recently released the latest system in its 3D optical scanner portfolio, the BLAZE 600M. The solution is similar to structured light but comprised of a blended combination of technologies that allow it to be faster and more accurate than traditional structured-light systems.

An explanation of how the systems works begins with how it reconstructs point cloud data using built-in, multiple-image acquisition modes.

Structured light 101

Structured light, or phase shift imaging, generally includes capturing successive images of patterns, typically stripes, projected onto an object. These projections are of known pattern, as is the shift between them. In order to identify each specific stripe in the projection, they are light-coded. The most common coding technique is known as "gray code." Each projection includes a set of stripes that have half the widths compared to the stripes of the previous projection, as shown in figure 1 below. The stripes in each projection are accurately shifted from those of the previous projection by a known phase. Two or more cameras of known distance and angle to the projector's axis capture these projected patterns.

Because the patterns are distorted when they are projected against geometric shapes, it's possible to use triangulation from the cameras, along with what's known about the patterns and the camera's positions, to calculate the 3D shape of the object. This 3D reconstruction process requires a series of projections taken over time, about four seconds on average. Of course, in order to capture the entire surface of an object, this capture process is conducted multiple times while rotating the object to be measured, or by moving the structured light camera around the object.

Figure 1: An example of gray code

The speed factor

One key factor in speeding up this process is to limit the number of patterns projected, and limit the number of camera positions required to capture the entire surface of the object. The Blaze 600M uses only a single random pattern projection from a slide projection or digital light processing (DLP) pattern. The image acquisition process requires only milliseconds rather than the four or so seconds mentioned above and is used by Hexagon's 3D scanners today. In this system, rather than stripes, a random pattern is built from interchanging dark and bright contrast areas, as seen in figure 2 below:

Figure 2: Example of random pattern projection

The software matches each point in one camera with the corresponding point in view of the second and third cameras. This process is performed for all image points simultaneously. A dedicated algorithm engine calculates matching points to sub-pixel accuracy by using neighboring pixel information (also called spatial "correlation"). This process is achieved using triangulation, where the pattern is "compared" on a sub-pixel level using all three cameras. This method enables the system to rapidly determine and capture 3D points on the object surface, based on the correspondent points in the images and the camera's geometry calibration.

In most structured light systems, two cameras are used for identifying a point in its 3D position. The Blaze 600M uses a third camera to provide an additional point of reference and enhances the pixel matching described above. When using a single slide projection instead of multiple projections, the third camera plays a larger role in the reconstruction, which enables a high level of confidence in reconstruction along with fast image acquisition time.

The presence of the third camera also enables reconstruction of data collected by two cameras, and provides accessibility to areas that can't be seen by all three cameras. Therefore, although ideally it would be best if all three cameras could see a point, it is possible to reconstruct data if only two of the three cameras can see a particular point due to line-of-sight issues. This versatility means that more usable data can be gathered from any image acquisition step, potentially decreasing the number of shots required to capture the entire surface of the object.

The BLAZE 600M also incorporates a programmable DLP technology into the solution that enables the projection of various patterns, in addition to a random pattern. The more patterns you have to work with, the more you can build better point-cloud data out of the sets of images. By better, we mean less noise and improved geometry tracking. The DLP unit offers various structured-light modes, enabling users to obtain multiple images taken with different patterns while keeping acquisition times relatively short. The advantage of the DLP is that it is programmable, and there are no moving parts; some 3D optical scanners use mechanical slides for the shift, which increases acquisition time.

By including a third camera and leveraging an enhanced form of structured light technology, we're able to meet production process demands for high throughput in challenging conditions. Examples include press-line operations and multiple operations performed simultaneously in the aerospace industry.

Hybrid 3D acquisition

When both technologies are merged under a single sensor platform, hybrid 3D acquisition modes can be generated. These hybrid modes can benefit from the advantages of both methodologies to create unique acquisition modes. The advance structured light (ASL) mode is one example of using the random pattern projection with dedicated structured light patterns, as shown in figure 3 below.

Figure 3: ASL single-pattern projection example

Using a combination of these patterns, along with burst-image acquisition mode, the system can gather more information on each individual pixel with a very short image-acquisition time, averaging 40 milliseconds. The additional information is processed by a surface reconstruction engine to improve point-cloud noise, in terms of density and coverage, as well as point-cloud quality on low-reflective (e.g., black) surfaces, highly reflective surfaces, and non-uniform surfaces. The acquisition of finer details on the measured object is also improved with the additional data.

As an example, the STL data shown in figure 4 was obtained from a highly reflective, black-painted surface without preparing the surface with spray or powder. The left image shows data acquired using a common structured light sensor. The right image shows the same surface captured with the BLAZE 600 using ASL mode. The right image is based on a combination of single slide projection and additional pattern such as the checkered pattern.

Figure 4: The BLAZE 600M data (right) compared to that from a typical optical scanner (left)

The actual scanned part is shown below in figure 5. The part is painted black and has a very shiny finish.

Figure 5: Actual part, with blue light scanning the surface

Accuracy enhanced

The BLAZE 600M's ability to acquire rich and high-quality 3D point cloud data is only part of the accuracy equation for this 3D optical scanner. When a scanner measures closed features on a part's surface, such as holes, slots, or edge areas (trimmed or hemmed), point cloud data as a single source of measurement may not provide the most accurate results. When points on trim edges or holes are reconstructed, the reconstructed point cloud may be noisy or incomplete with respect to the actual edge of the feature. Therefore, the BLAZE 600M uses dedicated images for edge detection, focusing on the contrast between the part and the background. This isn't a single projection but a different image taken as part of the image-acquisition process for feature measurement using unique, 2D image analysis.

As shown in figure 6, the first image on the top left contains the pattern projection and point-cloud reconstruction. The image on the bottom left is a different image. The feature image does not have the pattern projection and is used to detect the edges of the measured feature based on contrast. By using edge detection and image processing engines, the feature edge can be accurately located in its 2D position. Combined with the point cloud data, the feature's 3D result can be calculated.

Using 2D image analysis methodology, users can leverage best building practices to reduce variances in measurement and ensure reliable data from scanning parts or assemblies.

Figure 6: The combination of random pattern, point cloud, and 2D feature images results in very accurate part scanning

Factory vibration? No problem.

As stated, the BLAZE 600M provides image acquisition in milliseconds and enables operators to take more images in less time. In addition to reducing measurement time, a major benefit of the optical scanner is its ability to operate under harsh shop-floor conditions. A good example is an automotive stamping facility where press lines are generating a high-vibration environment. Heavy cranes move dies around the floor and change the ambient illumination at any point. Due to the optical scanner's ultra-fast image-acquisition process, vibration and light challenges are mitigated.

In a high-vibration environment, the scanner position relative to the part being inspected can change during the process. If a traditional scanning system takes multiple seconds to capture an image in this setting, it's inevitable that the movement of the scanner will prevent proper reconstruction of the image. Using the BLAZE 600M, the images are captured in milliseconds so physical movement of the scanner is unnoticeable, and measurement data can be reliably obtained in these conditions. The scanner's quick image attribution and specialized features maintain high data integrity on the factory floor. This information has a valuable impact on operations, since measurements can be performed without transporting the part. Proactive action can also be taken to reduce scrap and ensure parts are meeting specifications.

Shop-floor demands

Today's advanced manufacturers require agile solutions. For example, an automotive supplier must measure shiny sheet-metal parts in a robust production environment. Strong vibrations from operating presses are the norm for this facility. Nonetheless, the manufacturer needs to perform die development initiatives by measuring highly reflective operational panels on the shop floor and deliver functional builds. The BLAZE system is designed to excel in demanding, real-world scenarios.

Aerospace manufacturers face different challenges than automotive manufacturers. Although the aerospace production environment is typically calmer, workers are performing measurement tasks and conducting manufacturing activities on the same assembly. For example, an aircraft wing skin must be measured and at the same time, engineers need to work on the other side of the wing. Because multiple tasks take place simultaneously, the activity generates movement in the assembly. In this situation, a fast, vibration-resistant 3D optical scanner can be used to measure part quality and obtain actionable information without interrupting the manufacturing line during the course of a typical workday.

The scanner's effective and versatile use for in-place measurement of parts and assemblies is a step forward in the metrology world. Technicians use the BLAZE system to perform root cause analysis, enabling successful die tryouts, closure measurement, and virtual assembly builds. With reliable data, technicians can identify production issues, obtain highly accurate part inspection, and enhance the overall manufacturing process at a new level.

Figure 7: BLAZE 600M


About The Author

Amir Grinboim’s picture

Amir Grinboim

Amir Grinboim is the WLS product manager at Hexagon Manufacturing Intelligence. Grinboim has deep technical and engineering experience in the metrology space, specifically with white and blue light scanning systems. His international business experience includes sheet metal manufacturing with automotive, aviation, and industrial molding applications. Grinboim holds a degree from The Interdisciplinary Center in business administration and information technologies.