Figure - uploaded by Tomislav Medic
Content may be subject to copyright.
Source publication
Terrestrial laser scanner measurements suffer from systematic errors due to internal misalignments. The magnitude of the resulting errors in the point cloud in many cases exceeds the magnitude of random errors. Hence, the task of calibrating a laser scanner is important for applications with high accuracy demands. This paper primarily addresses the...
Similar publications
Terrestrial laser scanner (TLS) measurements suffer from systematic errors due to internal misalignments. The magnitude of the resulting errors in the point cloud exceeds the magnitude of random errors in many applications. Hence, the task of calibration is important for using laser scanners at applications with high demands regarding accuracy. How...
Citations
... Without any preregistration, the recorded and exported points of a scanning station refer to the local reference frame of the respective station. The scanned points can be converted from the Catesian form [x, y, z] to polar observations [r, , ] using known trigonometric functions [78]. ...
Laser scanning is a wide-spread practice to capture the environment. Besides the fields of robotics and self-driving cars, it has been applied in the field of engineering geodesy for documentation and monitoring purposes for many years. The registration of scans is still one of the main sources of uncertainty in the final point cloud. This paper presents a new keypoint-based method for terrestrial laser scan (TLS) registration for high-accuracy applications. Based on detected 2D-keypoints, we introduce a new statistical matching approach that tests wheter keypoints, scanned from two scan stations, can be assumed to be identical. This approach avoids the use of keypoint descriptors for matching and also handles wide distances between different scanner stations. The presented approach requires a good coarse registration as initial input, which can be achieved for example by artificial laser scanning targets. By means of two evaluation data sets, we show that our keypoint-based registration leads to the smallest loop closure error when traversing several stations compared to target-based and ICP registrations. Due to the high number of observations compared to the target-based registration, the reliability of the our keypoint-based registration can be increased significantly and the precision of the registration can be increased by about 25 % on average.
... This exemplary emphasizes that the stochastic model is not just constructed by the variances of the single observations, but the scanning procedure as a whole must be considered. Strategies exist to reduce the influence of systematic errors as some influences, such as internal misalignments, can be calibrated, e.g., [Holst et al., 2016a;Abbas et al., 2014;Chow et al., 2011;Lichti, 2007;Medić et al., 2017;Reshetyuk, 2009]. However, they can only be removed up to a certain precision, and some of the calibration parameters were found to change temporally . ...
... Based on these 18 parameters, Medić et al. [2020] established a user-oriented, efficient calibration field at Campus Klein-Altendorf (University of Bonn), which will be used in the further course of the thesis. Furthermore, they reduced the 18 parameters to ten relevant parameters for high-end panoramic-type scanners [Medić et al., 2017]. ...
Terrestrial laser scanners (TLS) are suitable for the surface approximation of objects and their geometric changes due to the temporally and spatially high-frequent data acquisition. However, precise geodetic engineering tasks require detailed knowledge about the performance of the sensors and especially about their uncertainty to use them for precise measurements, e.g., in deformation analysis or surface approximations.
Due to the complex transition behavior between error sources and effects on the point cloud, the correct description of the point cloud's stochastic model represented by the variance-covariance matrix is not yet solved. In particular, the interaction of the laser beam with the environment and the measurement object, taking into account different measurement arrangements (distances and angles of incidence), is so diverse that it is impossible to model all errors. However, if these errors are neglected in the stochastic model, this can lead to biased surface approximations, incorrect statistical tests, or misinterpreting errors as deformations. For this reason, strategies for the empirical determination of the stochastic properties of terrestrial laser scans are developed in this dissertation. In particular, the determination of the range precision for different surfaces and measurement configurations, as well as correlations between individual measurement points, are in focus. Specifically, the following aspects are addressed:
The object surface and scanning configuration mainly influence the range precision, which the reflected intensity of the laser beam can fully describe. This work contributes to efficiently determining the range precision by presenting a test field simplified for users and further developing the existing methodology. This contributes to a more realistic description of the main diagonal of the variance-covariance matrix representing the stochastic model.
Especially the interaction of the laser beam with the object is individual as it depends on the surface. The laser spot is integrated over a certain area, and neighboring laser spots overlap due to the dense acquisition of data points. This results in a smoothing effect and leads to the fact that the resolution capability of the scanner does not match the resolution set in the scanner. This thesis develops a new method for determining the resolution capability, which enables a more economic measurement planning. Furthermore, correlations are derived from overlapping laser spots, which are integrated into the stochastic model.
These rather short-scale correlations can be determined empirically via another method developed in this thesis. For this purpose, the stochastic signal of the point cloud must first be separated from the deterministic part. This is done with the help of a reference geometry generated with a sensor of higher accuracy. Subsequently, this work presents two methods for quantifying the short-scale correlations in the point cloud.
The previous methods can be implemented well for point clouds of smaller objects (approximately up to 2 m x 2 m). However, this is not straightforward to realize for larger objects as the stochastic properties change within the point cloud. Furthermore, a reference geometry is not easy to establish due to a lack of suitable sensors and deformations of the reference objects. For this reason, this thesis presents a method for creating a reference geometry of a larger object that allows for the analysis of long-scale correlations.
These different aspects provide a better understanding of the uncertainties in terrestrial laser scanning and, thus, form the basis for setting up a more realistic stochastic model of the point cloud to make statistically more reliable statements in a deformation analysis and unbiased surface approximations. Furthermore, the presented strategies do not require special laboratory conditions but can be performed by qualified users if an appropriate object, such as a roughly planar wall, is available.
... Not all 18 parameters are determinable through typical calibration routines. For this reason, a simplified version of this model has been adapted by Medić et al. (2017) and used for high-end panoramic scanners. From the 18 CPs, they determine 10 as relevant (see tab. 3.2) in reducing most of the systematic instrumental errors, but thorough this thesis, only the random component of the CPs is of interest. ...
... From the 18 CPs, they determine 10 as relevant (see tab. 3.2) in reducing most of the systematic instrumental errors, but thorough this thesis, only the random component of the CPs is of interest. , Medić et al., 2017. Besides the graphical representation of the CPs, more detailed explanations are not given here, since they have been extensively discussed by Muralikrishnan et al. (2015) and Medić et al. (2017). ...
... , Medić et al., 2017. Besides the graphical representation of the CPs, more detailed explanations are not given here, since they have been extensively discussed by Muralikrishnan et al. (2015) and Medić et al. (2017). The following figures (inspired by Muralikrishnan et al., 2015;Medić et al., 2017;Muralikrishnan, 2021) illustrate the CPs separately for clearness. ...
This work presents a method to define a stochastic model in form of a synthetic variance-covariance matrix (SVCM) for TLS observations. It relies on the elementary error theory defined by Bessel and Hagen at the beginning of the 19th century and adapted for geodetic observations by Pelzer and Schwieger at the end of the 20th century. According to this theory, different types of errors that affect TLS measurements are classified into three groups: non-correlating, functional correlating, and stochastic correlating errors.
For each group, different types of errors are studied based on the error sources that affect TLS observations. These types are classified as instrument-specific errors, environment-related errors, and object surface-related errors. Regarding instrument errors, calibration models for high-end laser scanners are studied. For the propagation medium of TLS observations, the effects of air temperature, air pressure and vertical temperature gradient on TLS distances and vertical angles are studied. An approach based on time series theory is used for extracting the spatial correlations between observation lines. For the object’s surface properties, the effect of surface roughness and reflectivity on the distance measurement is considered. Both parameters affect the variances and covariances in the stochastic model. For each of the error types, examples based on own research or literature are given.
After establishing the model, four different study cases are used to exemplify the utility of a fully populated SVCM. The scenarios include real objects measured under laboratory and field conditions and simulated objects. The first example outlines the results from the SVCM based on a simulated wall with an analysis of the variance and covariance contribution. In the second study case, the role of the SVCM in a sphere adjustment is highlighted. A third study case presents a deformation analysis of a wooden tower. Finally, the fourth example shows how to derive an optimal TLS station point based on the SVCM trace.
All in all, this thesis brings a contribution by defining a new stochastic model based on the elementary error theory in the form a SVCM for TLS measurements. It may be used for purposes such as analysis of error magnitude on scanned objects, adjustment of surfaces, or finding an optimal TLS station point position with regard to predefined criteria.
... Moreover, procedures are shown for determining the parameters of these models using either sys-1. Introduction tem calibration or individual component calibration, Lichti and Franke (2005), Rietdorf (2005), Lichti and Licht (2006), Lichti (2007), Reshetyuk (2009), Gordon (2008, Zogg (2008), Medić et al. (2017), Holst et al. (2018), Lichti et al. (2000), Medić (2021). Some challenges still prevail in scanner calibration, and companies and research institutes are advancing the calibration methods. ...
Reflectorless electronic distance measurement (RL-EDM) based on optical signals, transmitted from the instrument and reflected by natural surfaces, enables fast, accurate 3d mapping and digitization of the environment using terrestrial laser scanning (TLS). The measurements are naturally affected by the instrumental imperfections, atmospheric effects, and surface and material effects that primarily involve the geometry and physical properties, roughness, penetration, and anisotropic reflections of the scanned surfaces. This consequently limits the achievable accuracy or makes it challenging to make reliable accuracy predictions for applications where this is needed. In practice, the surface-related effects are the most influential ones for scanning distance of tens to hundreds of meters.
In this thesis, a numeric model for the light detection and ranging (LiDAR) measurement process has been developed and implemented for numerical simulations to understand and study the surface-related effects and their influence on the accuracy of the TLS measurements. The numeric simulations follow a geometrical optics approach, assuming a fundamental Gaussian laser beam profile. The measurement process involves a continuous wave (CW) phase-based RL-EDM with I/Q-demodulation or a pulsed wave (PW) based RL-EDM with pulse detection technique. The main underlying assumptions involve the beam parameters, the surface topography within the measurement footprint, the material properties within the measurement footprint, the geometric configuration, which involves the distance and angle of incidence, noise at the detector, and the measurement principle. The simulations are extended from RL-EDM to a 3d laser scanning process by deflecting the measurement beam into incrementally changed spatial directions.
Numeric simulations are used herein to investigate the effects of the angle of incidence, surface curvature, mixed pixels, and surface roughness, approximating the surface geometry by a triangular irregular network (TIN) of high spatial resolution. The surfaces are assumed to be perfectly diffuse; an absolute reflectance and a Lambertian scattering model are allocated to each triangle in the network. The LiDAR equation is implemented to compute the power received at the detection unit.
The simulation outcomes are compared to the corresponding experimental results obtained with scanning experiments using phase-based laser scanners (mostly Z+F Imager 5016 and Faro X330) both indoors and outdoors. In addition, experimental investigations of a few selected surface specimens, namely, spruce wood, beech wood, and concrete using a Z+F 5016 scanner, are carried out to understand the impact of the underlying surface-related effects on the deviations and noise of the laser scanning measured points. The experimental studies help to understand the plausible reasons for the discrepancies between the simulated and real scan results. As an additional contribution, a simple procedure for retrieving and deriving sufficient approximations of the beam parameters of a phase-based laser scanner experimentally is developed. They are necessary for realistic simulations.
... The coefficients of the systematic error models, denoted herein as additional parameters (APs), can be estimated using special-purpose, self-calibration methods. A number of research teams have developed and successfully demonstrated TLS models and self-calibration methods (García-San-Miguel and Lerma, 2013;Li et al., 2018;Lichti, 2007;Medić et al., 2017;Muralikrishnan et al., 2015;Reshetyuk, 2010). Highcontrast, signalized targets are commonly used as the primitives for the self-calibration. ...
... The fourth term, c 2 , is included here owing to its dependence on elevation angle and because its existence has been recently reported for several instruments (Lichti et al., 2019a). It is recognized that more comprehensive error models have been proposed (Muralikrishnan et al., 2015) and utilized by others (Medić et al., 2017). The model used herein comprises APs that have been identified through sequential model identification. ...
Terrestrial laser scanning (TLS) is established as a viable means for precision measurement and the need for systematic error modelling and instrument self-calibration is well recognized. While additional parameter (AP) models and procedures for their estimation from signalized target fields have been developed, the first-order design (FOD) of TLS self-calibration networks remains an active area of research aiming to improve AP quality. The conventional FOD approach of numerical simulation carries a heavy computational burden. This paper reports a new method for TLS self-calibration FOD that avoids the high computational effort and can predict AP precision in closed form. Its basis is a relatively simple analytical model of the distribution of spherical coordinate observations, specifically the elevation angle. The accuracy of predicted AP precision is quantified by comparison of precision estimates from a more complex and detailed observation distribution model and from self-calibration. Results from 25 datasets demonstrate the high accuracy (arc second or better) of the closed-form approach. A new observation distribution model is then developed to optimize the geometric design of TLS self-calibration networks. An ideal observation distribution based on the versine function and a corresponding target field configuration that enhance AP precision are established. Testing was performed on five additional, very dense TLS self-calibration datasets. Each dataset was subsampled so as to replicate the observation distributions corresponding to conventional network design and the proposed design. The results show that up to 55% improvement in AP precision, obtained from self-calibration, can be achieved with the new design and these results agree with versine-distribution model predictions within 14–16%.
... This model was developed primarily to assist in designing test procedures that are sensitive to the different error sources in support of the development of documentary standards within ASTM. This model has since been adopted by other researchers [63,66,67]. We briefly address such sensitivity analysis based test-position determination in Section 5.2. ...
Terrestrial laser scanners (TLSs) are increasingly used in several applications such as reverse engineering, digital reconstruction of historical monuments, geodesy and surveying, deformation monitoring of structures, forensic crime scene preservation, manufacturing and assembly of engineering components, and architectural, engineering, and construction (AEC) applications. The tolerances required in these tasks range from few tens of millimeters (for example, in historical monument digitization) to few tens of micrometers (for example, in high precision manufacturing and assembly). With numerous TLS instrument manufacturers, each offering multiple models of TLSs with idiosyncratic specifications, it is a considerable challenge for users to compare instruments or evaluate their performance to determine if they meet specifications. As a result, considerable efforts have been made by research groups across the world to model TLS error sources and to develop specialized performance evaluation test procedures. In this paper, we review these efforts including recent work to develop documentary standards for TLS performance evaluation and discuss the role of these test procedures in establishing metrological traceability of TLS measurements.
... The two-face method has its origin in Neitzel (2006) where it is used for a simplified TLS calibration experiment. It is introduced in the TLS self-calibration in its current form within this thesis in publication A1 (Medić et al., 2017). An alternative form was developed shortly after in Wang et al. (2017). ...
... Solving the task of designing an efficient calibration field (FOD problem) requires clearly defined and accurate functional model of the calibration adjustment (Section 2.1.2). As this was not entirely fulfilled in the literature, in the publication A1 (Medić et al., 2017), this functional model is investigated and adapted for the target-based selfcalibration of the high-end panoramic TLSs as explained in Sections 3.1.1 and 3.1.3. ...
... Eq. 3.1-3.3) was implemented in the way that allows the explicit introduction of the two-face measurements, which were proven to be an indispensable source of information for the calibration in the publication A1 (Medić et al., 2017). ...
Terrestrial laser scanner (TLS) measurements are unavoidably affected by systematic errors due to internal mechanical misalignments, while the calibration is the fundamental procedure for mitigating these effects and assuring the high measurement accuracy. Despite different existing calibration solutions, the practice of calibrating TLSs for the sake of quality assurance is rarely exercised. The main reason is the inefficiency of the existing solutions requiring considerable investments of time, funding and effort. Therefore, this thesis is dedicated to increasing the efficiency of the existing calibration approaches, development of further efficient calibration approaches, and aiding the selection of the appropriate calibration strategy. It can be separated into three main aspects: First, an optimized calibration field design is derived for the well-established target-based self-calibration approach. It comprises only a handful of targets and requires typically one hour for the calibration procedure, while at the same time assures trustworthy and comprehensive calibration. Second, as the stability of the calibration parameters is found to be one of the limiting factors, the first fully automatic in-situ calibration strategy was introduced in the response to identified efficiency related shortages of the existing in-situ calibration strategies. Finally, within the analysis of different calibration approaches, it was proven that the user-calibration of instruments can indeed improve the point cloud accuracy beyond the manufacturer’s initial factory calibration. Moreover, it is proven that, if necessary, user-calibration can even substitute the manufacturer’s calibration entirely. In the overall view, the findings and methods developed in this thesis contribute to the integration of the TLS calibration practices in the commercial applications and aid their wider acceptance.
... Preliminary study to determine the parameters of the NIST error model for a TLS by selfcalibration was attempted by Hughes et al. [24] through measurements of 26 targets at four TLS positions and Medic et al [25] through measurements of 291 points at only one TLS position. ...
... Numerous published papers on this topic, see refs [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25]. Our contribution here is that we have considered a more comprehensive error model, see ref [22][23]. ...
In this paper, we discuss two aspects concerning terrestrial laser scanners (TLSs) – error model calibration and performance evaluation. Error model calibration is the process of determining parameters of an error model to improve the accuracy of TLSs. Performance evaluation refers to a series of tests to determine if a TLS meets specifications provided by the manufacturer. Both procedures can be accomplished using a network of stationary targets whose locations are known from a prior calibration using another method/instrument. This paper explores the question of whether the network (i.e., target locations) must be calibrated using an instrument of higher accuracy such as a laser tracker (LT) or whether the TLS under study is itself suitable for network calibration. Regardless of whether an LT or a TLS is used, the calibration is performed from target measurements made from multiple locations of the instrument to average out systematic errors and reduce the uncertainties in target coordinates. Such approach of multi-position measurements on stationary targets is referred to as the network method. We provide guidance on when the TLS is sufficient for network calibration and when an LT may be necessary for performance evaluation purposes.
... Mobile machine vision is a kind of portable visual image processing method, on the road traffic, the traffic in the test the traffic state identification. Mobile machine vision can make up for the defect of traditional inspection methods, does not depend on capital equipment, can be flexible to regional testing and monitoring, reduce the amount of monitoring equipment, greatly reduces the cost of equipment, is a supplement to perfect the existing traffic system inspection method of [9][10]. ...
Research on computer vision technology based on deep learning has become an important research direction in the field of artificial intelligence. Among them, the target detection task of images and videos has occupied a pivotal position in many intelligent vision research and applications. The purpose of this article is to learn from the existing object recognition and target detection technology foundation, to conduct in-depth research on the current traffic patrol deficiencies, to achieve real-time patrol and monitoring of key roads. This article chooses to use the target detection algorithm of Faster R-CNN in deep learning, applies drones and mobile machine vision to traffic patrol inspection, and detects various targets in the PASCAL VOC data set. The AP of the bus is 68.20%; The AP of the car is 75.50%; The moving object recognition and tracking method based on Faster R-CNN and Hungarian matching method is verified, and traffic flow data can be calculated to realize real-time monitoring of road traffic. The research in this paper helps to overcome the defects of human eye recognition in the traditional monitoring system and improve the efficiency of traffic inspection.
... Illustration of Panoramic Terrestrial Laser Scanner Geometry(Medic, et al., 2017)................................................................................................................................................... Working Principle of Total Station Distance Measurement Using Reflector Prism (Afeni, 2012)………………………………………………………………………………… 66 Figure 4.2: Description of a Leica Total Station TCR 1201+ (Afeni, 2012) ........................... 67 xi Standard Leica GPR1 Circular Prism .................................................................... Closed Traverse Route to Calculate the Coordinate of the Control Beacon ......... Principle of Levelling and an Automatic Level .................................................... Image of the Control Beacon in the Laboratory .................................................... Illustration of Force Centring Beacon Design (Thomas, 2011) ............................ Plan View of the Scanning Laboratory ................................................................. 10: Description of Asymmetric Target Placement Plan ............................................ 11: Photograph Showing Placement of Trimble Checkerboard Targets in the Laboratory ................................................................................................................................ 12: An Image Showing the Placement of Targets on the Concrete Wall and the Ceiling of the Scanning Laboratory ...................................................................................................... 13: Photograph of the Terrestrial Laser Scanner Placed on the Control Beacon and Tripod Stand............................................................................................................................. Faro Focus 3D, Picture Courtesy of Faro………………………………………..High Resolution Spherical Image of the Scanning Laboratory with full Colour Image........................................................................................................................................ Theoretical Image of Time of Flight (Kaldén & Sternå, 2015)............................. Illustration of ICP Distance Calculation for Conjugate Points (Vosselman & Mass, 2010). ....................................................................................................................................... Illustration Showing the Integration of Point, Line, and Plane Features for Registering LiDAR Point Cloud Data (Jaw & Chuang, 2008). ...
... The size of the test facility is crucial in determining the number of setup and range offset. According toMedic et al., (2017), a bigger facility increases the quality of the testing process. The availability of long range and large space reduces the angular correlations between parameters, thereby causing an increase in reflected laser signal from the targets.Most of the parameters used in testing the instrument are angular based variables which depend on the scanner orientation angle. ...
... Illustration of Panoramic Terrestrial Laser Scanner Geometry(Medic, et al., 2017). ...
Laser Scanning is a 21st century surveying technique used to generate high density and point cloud data in 3D for surveying, mapping and monitoring purposes, for example rock mass movement in mining. For high precision and accurate work, the instrument must be used correctly and with regular calibration and checks to ensure that it constantly performs according to expectations and manufacturer specifications. The aim of this research was to develop a short-range scanning laboratory for testing the accuracy of terrestrial laser scanning systems for rock engineering applications. This research was based on methods used by previous researchers in testing the accuracy of the instrument and the development of a suitable facility for such testing. The procedure used in developing the short-range testing facility included selecting an appropriate venue of size and shape that suits the requirements for a short-range laser scanning laboratory. This was followed by the construction of the master control beacon and creation of additional scan set ups in order to capture all the points in the facility. Targets were strategically placed on the wall and roof of the laboratory in order to determine the centre point coordinate of each target. A Leica Total Station TCR 1201+ and Trimble S6 Total Station were used to establish accurate coordinates for the control beacon and the targets respectively. Thereafter, the scanning of the targets was carried out using a FARO Focus XD 130 terrestrial laser scanner. Comparisons were performed using the coordinates from the terrestrial laser scanning and those of the Total Station to examine the point accuracy of the scans. The results from the comparison between the scanner coordinates and the total station coordinates showed that the FARO Focus laser scanner performed within manufacturer specifications but not always. This implies the instrument is capable of generating accurate point and reliable cloud data that can be used for the purpose of monitoring underground rock mass movements. Errors regarded mistakes in the final analysis occurred as a result of target design (in terms of size and orientations) and oblique lines of sight.
