Terrestrial Laser Scanning

Calibration of Terrestrial Laser Scanners

Terrestrial Laser Scanners have become a standard sensor extensively utilized in engineering geodesy and related sub-disciplines. Owing to continual advancements, its usage has expanded to high-quality measurement tasks. Nevertheless, internal misalignments introduce systematic deviations in the point cloud. Effective calibration strategies minimize these deviations.

A dedicated calibration field has been designed specifically for calibrating Terrestrial Laser Scanners, enabling the determination of essential calibration parameters and their stochastic information for high-end scanners with minimal effort (a-priori calibration). Considering these parameters significantly enhances the quality of resultant point clouds. However, certain crucial parameters for low-cost laser scanners cannot be ascertained in this calibration field. Additionally, some parameters exhibit temporal instability and substantial dependence on external conditions (e.g. temperature). In-situ calibration serves as a solution, where calibration parameters are directly determined based on the point clouds from the measurement task. This approach allows the determination of parameters relevant to low-cost terrestrial laser scanners that are directly applicable to the specific measurement environment.

Contacts:

• M.Sc. Eike Koller
• Dr.-Ing. Berit Jost
• Prof. Dr. Heiner Kuhlmann

Stochastic model of terrestrial laser scans

Using terrestrial laser scanners (TLS) to acquire geometric properties of surfaces has been proven to be a solid procedure in tasks of industrial and classical survey. New TLS measure with a rate of up to two million points per second and generate a highly dense three-dimensional point cloud. Accordingly, they have a high spatio-temporal resolution.

In terms of deformation analyses, the results depend on the stochastic characteristics of the point cloud. They are defined by the covariance matrix of the observations, which consists of variances and covariances. The variances of the range observations can be derived from a functional model as they solely depend on the intensity of the measured points. Yet, the correlations and, thus, the covariances are not sufficiently known and therefore neglected.

We develop methodologies to quantify covariances empirically to set up a fully populated covariance matrix of the point cloud. Therefore, we established a test field called “The Bonn Reference Wall” that has a known geometry to analyze point clouds for their uncertainty.

Contacts:

• Dr.-Ing. Berit Jost
• Prof. Dr. Heiner Kuhlmann

Point-based and Area-based Deformation Analyses

Since several years, the areal acquisition of objects and their geometries - up to an integrated space continualization - got into focus of engineering geodesy. Consequently, objects are no longer sampled and discretized by a small number of individually defined signalized points. Instead, they are area-wise sampled by a large number of non-signalized points whose individual positions cannot be set up. This fact makes completely new demands regarding the subsequent data processing: Up to now, deformation analyses are focused on analyzing single points or point differences to decide whether the sampled object is deformed at selected positions.

Since there are no longer signalized points when using laser scanners and, thus, also no longer identical points between two epochs, judging an object to be deformed can only be based on an areal parameterization. In the figure below, we compare two epochs of the Brucher water dam measured with TLS. The geometry of the dam changes up to 1 cm between summer and winter. However, the movement cannot be judged with a real, statistically correct deformation analysis, as the stochastic model of the point cloud is not yet sufficiently known.

Contacts:

• Dr.-Ing. Berit Jost
• Prof. Dr. Heiner Kuhlmann

Resolution Capability

The fast and dense data acquisition of terrestrial laser scanners (TLSs) gives many opportunities to sample objects with many details. Nowadays, TLSs are capable to sample up to two million points per second with a point spacing of less than a millimeter on ten meters, and a point accuracy of a few millimeters or even less. However, the size of objects that can be spatially resolved in a point cloud varies between different scanners as it depends on the point distance as well as on the laser spot size: If the object is scanned with a high resolution, neighboring laser spots overlap as the laser beam is at least a few millimeters. Hence, they do not provide individual information about the object and the resolution does not equal the resolution capability.

We develop methodologies to determine the resolution capability for different scanners to answer the following questions:

• Which objects can my scanner resolve?
• Which scanner can resolve my object?
• Which are the optimal settings for an economic and efficient data acquisition?

Contacts:

• Dr.-Ing. Berit Jost
• Prof. Dr. Heiner Kuhlmann

Integration of digital imagery into TLS point clouds

In recent years, the spatial capture of geometries using terrestrial laser scanners has moved into the focus of engineering geodesy. But especially at object discontinuities -particularly along edges- issues like mixed pixels, smoothing effects, missing the discontinuity etc. arises frequently. However, depending on the measurement project, the level of detail in the point clouds may not be sufficient.

Particularly at object discontinuities the integration of digital images into TLS point clouds provides the opportunity to enhance the level of detail. Due to the fusion of images and TLS point clouds it becomes possible to generate new points in the point cloud that represent object edges more accurately compared to the TLS point cloud itself. In this process, algorithms are strategically employed to maximize the complementary capabilities of the sensors to compensate their disadvantages.

TLS point clouds can be enhanced, for example, through the integration of UAV images. The resultant points exhibit a considerably reduced distance from the actual edge compared to points that can be identified in the TLS point cloud as “edge points”. The quality of the newly generated points depends mainly on the accuracy of intrinsic and extrinsic camera calibration and the Ground Sampling Distance (GSD) of the image.

Contacts:

• MSc. Eike Koller
• Dr.-Ing. Berit Jost
• PD Dr. rer. nat. Lasse Klingbeil
• Prof. Dr. Heiner Kuhlmann