Technology
Laser Triangulation
3D Scanner System Processes
Overview of Competing 3D Scanning Technologies
Choosing a Scanning Company
Laser triangulation is one of the most common techniques of 3D data acquisition. It is an active stereoscopic technique where the distance to the object is computed by means of a directional light source and a video camera. Figure One illustrates this triangulation principle.
A laser beam is deflected from a mirror onto a scanning object. The object scatters the light, which is then collected by a video camera located at a known triangulation distance from the laser. For facilitation purposes the camera is modeled as an ideal lens and the charged coupling device (CCD) detector is modeled as flat. It can be observed that the angle and the pixel position of the scattered light are related. Since the focal length of the camera lens is known, the analysis of the resulting video image can determine the angle of the scattered light.
The angle is also known since it is the projection angle of the laser beam. Thus using simple trigonometry the 3D spatial (XYZ) coordinates of a surface point can be determined. Hence the name triangulation.
There are several methods for detection and acquisition of the scattered light. The detector can be a one-dimensional (1D) or two-dimensional (2D) array of sensors. Today 2D arrays are common place as the need for faster and more accurate scanning techniques are demanded. As depicted in Figure Two a light line generator is used to highlight the object's surface.
The 2D array of the CCD camera captures the surface profile's image and digitizes all data points along the laser. A simple expansion of this technique allows the complete geometry of the object to be captured.
Laser
The laser employed by our Laser Scanning Systems produces a uniform intensity beam; typically, low-cost scanners exhibit non-uniform intensities (large variations in light sensitivity across different parts of the object). This lead to quality problems in the geometry capture process. Additional quality parameters are straightness, depth of focus, and mechanical stability of the laser assembly. Impact Studios' 3D scanners are also required to have very low scattered light, since any scattered light can result in stray signals from reflective parts of the object. The galvanometric scanning mechanism provides highly accurate and rapid positioning of the laser beam. Our scanners all use Class 1 laser devices, which are eye safe. Both hardware and software safety measures are built in, making the process completely safe.
Multi-line scanning
All of our systems make use of an advanced multi-line scanning process. This simple but powerful technique increases the scanning speed without sacrificing data (scan) integrity.
The scanners use a high-resolution NTSC black-and-white video camera for capturing geometry data. These cameras offer the most -effective solution for scanning. Unfortunately, there is a tradeoff. The NTSC format limits the camera's frame rate to 30 fps (or 60 fields per second). For instance, if a single laser line is projected per field, and the scan is acquired with 200 laser stripes, then the total scanning time exceeds 3 seconds. This is not acceptable for live objects. The multi-line process removes the NTSC data rate restrictions without effecting the quality of the scans, and scans in only one second.
Multi-line scanning is made possible through two innovations:
1) An ultra-fast scanning mechanism and
2) Proprietary software algorithms.The first allows the laser stripe's direction to change thousands of times per second. Since the stripe's positioning is now considerably faster than the data rate of the camera, multiple laser stripes may be acquired during a single video field. For example, with a multi-line factor of 4, the laser stripe's position may change at a rate of 1/240 seconds. This results in a projection of four laser stripes per video field (1/60 second). Thus, a multi-line factor of 4 is approximately four times faster than single line scanning. The second allows for the interpretation of the video field after scanning is complete. The complexity of these algorithms can be imagined in the following scenario. Extending the previous example with a multi-line factor of 4, suppose that some of the laser stripes do not hit continuously across the object. If only three of the stripes are captured interpreting the data becomes an arduous task. Even in this simple example there could be multiple interpretations of the data. Our proprietary algorithms allow the scanner to perform advanced geometry computations for topographically complex objects with any level of "multiplicity" factor. However, increased speed comes with a small price, that is, reduction in laser light intensity. Thus, while it is physically possible to achieve multiplicity factors higher than 10, it becomes impractical because the laser signal is reduced. Impact Studios generally scans live objects with a "multiplicity" factor of 4 or 5. In addition, we use a special software filter which compensates for object motion, thereby enhancing the multi-line process.
Geometry processing is accomplished using proprietary (FLF) Fast Laser Finder hardware located within our scanners. The FLF receives a digitized video signal from the black-and-white camera and performs real-time data processing to identify laser line positions.
Typically, the laser stripe spans approximately 6 pixels of the CCD along a single horizontal line. Several intensity thresholds and subtraction methods then allow the FLF to separate the laser stripe from the camera noise and background images. Finally, the FLF finds the center of mass of the laser stripe. With up to 15-bits of coordinate data, four of which are fractional, the calculations are performed with sub-pixel accuracy in real time. Also, remember that the FLF can process multiple scan lines per field, which is quite a few calculations!
Texture processing is accomplished using an advanced video grabber. The use of this device in conjunction with the FLF allows us to capture 3D data instantly while achieving the high fidelity graphics Impact Studios is known for.
Our scanners are not a coordinate measurement machine (CMM). The power of 3d laser scanning lies in its ability to measure thousands of point coordinates and create a 3D geometric clone of the scanning object instantaneously. A single scan typically generates more than 100,000 reproducible measurements (polygons).
Although triangulation is the most precise method of 3D, it has a common limitation;
increasing the accuracy increases the triangulation distance. Once again, there is a tradeoff; the larger the triangulation distance, the more shadows appear on the scanning object and the scanning head must be made larger. Our scanners are optimized to provide clear scans from as close as 20 centimeters up to 2 meters. The optimal scanning distance falls between 50 to 70 centimeters.
The dynamic range of our scanners is enhanced by its ability to change the scan multiplicity and adjust the aperture of the geometry camera. Thus, the objects as large as 2 m x 2 m can be scanned.
The accuracy of a 3D scanner is difficult to express in a single parameter. The accuracy depends not only on the surface and shape properties of the scanning object, but also the post-processing algorithms used to create data useful in 3D applications. In general, we can resolve features as small as 0.5 mm in any of the X, Y, and Z coordinate planes.
Using the Right ToolsetImpact Studios' powerful capabilities stem from our strong synergy of hardware and software. The software we use, contains a diverse tool. Key features include data filtering, dynamic polygon reduction, and scan fusion.
Data filtering is used to ensure 3D models are of the highest quality. Upon completion of scanning, the XYZ data point cloud is triangulated. The resulting number of triangles is approximately twice the number of data points. To minimize the scanner's statistical noise, a smoothing filter may be applied in which the points are placed in the center of mass of their neighbors. The points which are farthest from their neighbors are considered spikes, and are subsequently removed. Spikes can often be attributed to unwanted laser reflections from objects with highly reflective surfaces.
Capturing scans with upwards of 100,000 data points often become too large to work with on a personal computer and is well above the polygon budgets of most commercial applications. Impact Studios is able to relieve this data overload by employing an advanced algorithm for Dynamic Polygon Reduction. This algorithm, enables us to remove polygons (triangles) from a model in real time. Data is removed from the original model in an ordered fashion, removing the least important details first. All removed points remain in a "history" file, which allow the model to be "scaled" up and down indefinitely. The benefit of having Dynamic Polygon Reduction is that the model's geometry can be reduced while maintaining the visual integrity of the original model. We have seen models reduced from 80,000 polygons to 800 without any visual anomalies; that's a factor of 100! Reductions on this order of magnitude can only be achieved by starting with a high quality scan.
Laser scans are single projection (one viewpoint) only. Thus, for 360-degree model creation, a scan fusion (gluing) algorithm is critical. Once activated, the fusion algorithm finds the best rotation and translation matrices of the models, "glues" them together creating a single geometry for the object, and finally, re-triangulates. The result is a new seamless geometric model with properly blended textures. This approach permits gluing of almost any object with distinguishing characteristics in either texture or geometry. For example, a symmetric object with simple geometry but rich texture, would be a problem for most scanning services, but not for us. In addition, the new model may also be modified using our Dynamic Polygon Reduction Algorithm.
Overview of Competing 3D Scanning Technologies
This section will evaluate several competing technologies and explain the reasoning behind Impact Studios choice of scanning systems.
Stereoscopic vision is a passive optical technique. The basic idea is that two or more digital images are taken from known locations. The images are then processed to find the correlations between them. As soon as matching points are identified, the geometry can be computed. In spite of its simplicity (a minimum of two cameras), commercial systems based on this principle tend to be quite complex. Often involving more than two cameras, correlating points can not be repeatedly found for a wide range of objects necessitating the use of a special light projector. Furthermore, the processing time for a mid-resolution model is usually greater than ten minutes! Even when using video quality CCD cameras the resulting model's quality is considerably inferior to those obtained with laser-based systems. In order to obtain quality comparable to what Impact Studios provides, the camera's pixel size must be dramatically increased leading to higher system costs and longer processing times.
Moire projection triangulation scanners can offer a scan quality comparable to or greater than laser digitizers. However, systems achieving this quality sacrifice both speed and cost. In addition, the scan quality depends on material color properties, and large errors result when scanning a colored object. For example, scan a flat page from a colored catalog and you will see how rough the surface is. We believe that Moire scanners with comparable accuracy are considerably more expensive and less flexible than laser scanning systems.
Time-of-flight (laser range-finding) scanners provide a practical method acquiring large objects (e.g. buildings) with high-accuracy. Their main advantage is that scanning accuracy is constant across a wide range regardless of the distance to the object (in triangulation, the accuracy decreases linearly with the distance from the camera). However, the scanning speed is considerably slower (several readings per second) than triangulation scanners (the data rate of our scanners can be as high as 150,000 points per second). These scanners are also large and do not capture the objects texture, only its geometry. Currently time-of-flight scanners are not practical for fast digitization of small and medium-sized objects.
There are several laser triangulation scanners whose laser stripe position is fixed with respect to the camera (our scanners' laser stripe is moved by a galvanometric motor). In these scanners it is assumed that the scanner head or the object will be moved with respect to each other. Usually the scanner head is moved via a robotic arm either manually or automatically. Even though the viewing angle of such systems is larger than that of single-projection scanners, the acquired geometry must still be "glued" together. Thus, the advantage of a larger viewing angle is negated due to our Scan Fusion Algorithm.
Choosing a Scanning Company
Impact Studios invites you to use the following checklist when evaluating a 3D scanning service. We hope this list will enlighten you to the numerous advantages provided by Impact Studios.
 |
 |
 |
 |
 |
 |
 |
 |
