Accuracy of Measurement

Selective sections from GSA BIM Guide Series 03 – BIM Guide For 3D Imaging

3.1.1. Sources of Error
All measurements contain errors. Even measurements from a calibrated instrument are subject to random fluctuations or “noise”. Besides random errors, systematic errors will cause incorrect measurements unless corrected (e.g., applying a correction factor to the measurements). These errors, systematic and random, can originate from the instrument, operator, and/or environmental conditions. The processes that can introduce errors to the 3D imaging measurements or to the end-product of 3D imaging data are briefly described in the following sections.

3.1.1.1. Calibration
Errors from an improperly calibrated instrument are systematic errors and result in an offset or bias in the measurements. This offset may be constant, linear, nonlinear, or periodic …

For 3D imaging systems that acquire color information in addition to 3D data, calibration of the color information with the 3D data is also required.

3.1.1.2. Scanning
Field errors can arise from instrument error (3D imaging systems and traditional survey instruments), environmental conditions, and operator error. Instrument error has a systematic component (calibration error – see section 3.1.1.1) and a random component (instrument noise). Random errors include pointing error, centering error, leveling error, and reading error. …

Intrinsic to operator error is operator skill – operator of a survey instrument (if used) or the operator of a 3D imaging instrument. …

Measurement errors from scanning can come from a variety of sources. Measurements are affected by the scanned object’s surface characteristics. Most 3D imaging systems have problems measuring a highly reflective or specular object or surface (e.g., mirror, reflective material used for road signage, ice). More measurement noise and the higher likelihood of obtaining no measurement are associated with lower reflectivity (darker) surfaces and objects. Other surfaces that may result in no measurements include wet surfaces such as water puddles and wet asphalt. The material of the object also affects measurements – problems can arise when measuring glass, plastics, machined metals and marble. In the case of marble, some lasers penetrate the marble resulting in biased measurements.

Increased measurement noise also occurs when scanning objects at oblique angles. This noise increases as the angle increases.

Spurious measurements can also be obtained when scanning across edges where the laser beam is split by the edge (this effect is sometimes referred to as a mixed pixel), when scanning into the sun, range ambiguity (of phase-based systems), and by cross-talk (interference between signals in the electronics).
Environmental and ambient factors also contribute to scanning error. The thermal expansion of an object affects the measurements. Some examples are 1) scanning a pipe when it is hot and when it is cold and 2) scanning a wall heated by the afternoon sun and scanning the same wall at night. The measurements in these situations can be significantly different and can cause errors in registration and fitting. Heating of tripods can cause movement of instrument. Temperature gradients will also affect measurements. For example, an asphalt road heated by the summer sun will result in a temperature gradient near the road surface. Windy conditions can affect measurements by causing movement or vibration of the instrument or the structure on which the instrument is stationed. Rain, snow, dust, and moving cars and pedestrians when scanning roadways will also result in spurious and unwanted data.

3.1.1.2.1. Scan Plans
In general, the level of detail (resolution) that can be captured reduces with increasing distance from the 3D imaging system. This is because the point density reduces with distance, and the beam width increases with distance (beam divergence describes how the beam width increases as a function of distance and is generally included in the instrument specifications). For example, if the beam width were 6 mm x 6 mm at 30 m (0.25 in x 0.25 in at 100 ft), then identifying a feature size less than 6 mm (0.25 in) would not be feasible with the instrument at this location. The instrument would have to be located closer to the feature or some other method should be used to achieve the desired resolution. …

Resolution is also dependent on beam width, object reflectivity, angle of incidence with the object, object material/texture, scan speed (for scanning systems), and horizontal/vertical orientation of the feature relative to the beam direction.

3.1.1.3. Registration
Data registration is required to register two or more datasets (datasets from the same instrument obtained from different locations or datasets obtained from different instruments) so that they have a common coordinate frame (Figures 19 and 20) or to register a dataset to another coordinate frame (e.g., global, project). …

Currently, most registration is performed using targets. … The registration error using Method 1 is dependent on the how well the targets can be measured. In this case, the targets can
either be specific artifacts placed in the scene such as spheres and planar targets or recognizable and distinct features in a scene. …

surface-to-surface or surface-to-point cloud or point cloud to point cloud matching. … In general, this method
requires an initial registration where the scans are roughly aligned or a large region of overlap between the scans. As the measurement error is dependent on several factors (e.g., range, reflectivity, angle of incidence), the uncertainty of the points in the overlap region are unknown and using points with large uncertainties may not lead to an optimal registration.

A quantitative evaluation of these different registration methods is currently not available.

3.1.1.4. Modeling
Many 3D imaging applications require some kind of modeling — rendering of the scene in terms of surfaces, determining distances/volumes, or locating and identifying objects. The modeling process involves data editing (cleaning), segmentation, and fitting of geometric primitives (e.g., planes, spheres, cones, cylinders). Currently, most modeling is performed manually. It has been shown that an operator’s qualifications can have a large impact on the quality of a 3D model [2]. Therefore, the main sources of modeling errors are data editing, operator errors, and errors due to the choice of fitting algorithm. It should be noted that since the process of modeling introduces another error (modeling error), the uncertainty of the model is greater than the uncertainty of the raw data.
There are several sources of modeling errors. In some cases, models are simplified representations of the actual objects and this simplification may introduce errors. Depending on the application, this may or may not be acceptable. A second source of error can result by incorrectly inserting a duplicate model of a similar element. For example, all columns (Type A) on Floor 1 may have the same dimensions except for one (Type B) which was incorrectly constructed. In the modeling process, a model of Type A columns is created and duplicated for all columns on Floor 1 because the modeler incorrectly assumed that all columns were
identical. A third source of modeling error is the practice of measuring the inside dimensions of rooms and then using nominal wall thickness to represent voids. Based on construction techniques, actual wall thickness can vary dramatically.

It is important to remember that ALL measurements have errors. The reported measurement from the deliverables should be within the uncertainty of the secondary measurement system.

3.4. Personnel and training
There are currently no requirements in terms of licensing of 3D imaging service providers. Therefore, the experience of the service provider is critical. Service provider experience can be assessed by:
• work on previous projects of similar magnitude – for large projects, past experience not only speaks to the service provider’s ability to manage large projects but also the ability to handle large datasets
• accuracy achieved on previous projects
• training in the use of the hardware and software – familiarity with and understanding of the hardware and software are crucial for efficient work processes and in recognizing and resolving problems
• references from past projects
• an interview with the service provider – 3D imaging team and/or the post-processing team.

Post-processing of 3D imaging data and the generation of 2D drawings and 3D models from these data are time intensive tasks and require specific expertise. Determining a service provider’s experience in this area is as important as ascertaining his/her expertise in data collection.

For further information about this GSA BIM Guide Series 03 – BIM Guide For 3D Imaging visit the National 3D-4D-BIM webpage:  http://www.gsa.gov/graphics/pbs/GSA_BIM_Guide_Series_03.pdf