Get Latest Final Year Projects in your Email

Your Email ID:
PA Subs

Comparison of Roadway Roughness Derived from LIDAR and SFM 3D Point Clouds (Civil Project)

Download Project:

Fields with * are mandatory

ABSTRACT:

This report describes a short-term study undertaken to investigate the potential for using dense three-dimensional (3D) point clouds generated from light detection and ranging (LIDAR) and photogrammetry to assess roadway roughness.

Spatially continuous roughness maps have potential for the identification of localized roughness features, which would be a significant improvement over traditional profiling methods. This report specifically illustrates the use of terrestrial laser scanning (TLS) and photogrammetry using a process known as structure from motion (SFM) to acquire point clouds and illustrates the use of these point clouds in evaluating road roughness.

Five roadway sections were chosen for scanning and testing: three gravel road sections, one portland cement concrete (PCC) section, and one asphalt concrete (AC) section. To compare clouds obtained from terrestrial laser scanning and photogrammetry, the coordinates of the clouds for the same section on the same date were matched using open source computer code.

The research indicates that the technologies described are very promising for evaluating road roughness. The major advantage of both technologies is the large amount of data collected, which allows the evaluation of the full surface. Additional research is needed to further develop the use of dense 3D point clouds for roadway assessment.

 OVERVIEW OF ROUGHNESS EVALUATION TECHNIQUES

Pavement roughness has been a major pavement performance indicator since the early twentieth century (Buchanan et al. 1941,Nakamura 1962). Gillespie et al. (1980) described  methods to quantify pavement roughness using time-stable calibrated response
type systems.

After 1986, the international roughness index (IRI) was established as a standard pavement roughness measurement (Sayers et al. 1986a,Sayers et al., 1986b). IRI and other roughness indices are summary statistics, and since their development the main focus has been on increasing the accuracy of measurement and the repeatability
of results.

However,there is little agreement on how to quantify local feature roughness, even though studies have shown that local features are crucial to human comfort and cause most vehicle fatigue damage (Steinwolf et al. 2002, Oijer and Edlund 2004,zogsjö and Rychlik 2009).

DATA COLLECTION METHODS

Two technologies were chosen to acquire 3D point clouds of the road sections under consideration: (a) stationary or terrestrial 3D laser scanning (LIDAR) and (b) photogrammetry.

Both technologies produce 3D point clouds, where each point has x, y, and z coordinates, and thus both technologies can provide measurements in a 3D space.Descriptions of both technologies and the procedures followed in collecting data are provided below.

Data Collection Using Stationary Laser Scanner:

LIDAR systems measure the information (spatial coordinates and color) of a 3D space and store the information in a 3D point cloud. The term LIDAR is generic and includes airborne laser scanning technologies, mobile scanners mounted on vehicles, and stationary laser scanners or stationary terrestrial laser scanners, where the laser scanner is fixed at a station with known coordinates and a geospatially referenced 3D point cloud is constructed based on the distance between the scanner and the detected points.

Data Collection Using Photogrammetry:

Photogrammetry is defined as “the art, science, and technology of obtaining reliable information about physical objects and the environment through processes of recording, measuring, and interpreting photographic images and patterns of electromagnetic radiant energy and other phenomena” (Slama 1980). Modern photogrammetry has been revolutionized by the process known as structure from motion (SFM).

Alternative Georeferencing Tools:

In addition to the traditional flat and spherical targets provided by the manufacturer,a 76.2mm diameter spherical target, was designed and printed at the Iowa State University College of Design for $40. The cost of Trimble targets previously purchased was more than $100 per sphere.

3D-Printed Roadway:

Another product of the study that serves as a visual aid was a 3D-printed model of a roadway. The 3D print was for a corrugated gravel section with evident rutting. The point cloud collected via photogrammetry was used to construct a mesh utilizing a Poisson algorithm in Blender software.
TEST SECTIONS AND EXPERIMENTAL TEST RESULTS

Five sections were chosen for scanning and testing: three gravel sections, one portland cement concrete (PCC) section, and one asphalt concrete (AC) section. Samples were collected from the gravel sections when possible to conduct sieve analyses.

However, no samples were collected for the Story County section because the surface consisted of a large-size crushed limestone layer over a stiff crust. Dynamic cone penetration (DCP) tests were conducted at three test points (TPs) for each lane. Based on the data collected and using the DCP index (DCPI), defined as the amount of penetration divided by the number of blows causing the penetration, California bearing ratios (CBR) were estimated using equations (1), (2),and (3):

for CBR>10, DCP-CBR = 292/(DCPIx25.4)1.12   (1)
for CBR>10, DCP-CBR = 292/(DCPIx25.4)1.12   (2)
for CBR>10, DCP-CBR = 292/(DCPIx25.4)1.12   (3)

 DATA ANALYSIS METHODS

After cleaning and segmenting the point clouds, the data were processed using algorithms developed to produce roughness maps. The first step in analyzing the data is to form a mesh grid from the discrete measurement points.

The grid elements use predefined x and y edge dimensions to form a grid region. The grid’s center elevation is calculated as the median of all cloud points falling within that grid region. All points are rotated and translated to a local coordinate system corresponding to the longitudinal and transverse axes (Alhsan et al. 2015).

Roughness Analysis:

The evaluation of roughness in this study is based on the responses of a mechanical system that approximates the response of a passenger vehicle. The mechanical system simulated here in is the quarter-car model described in ASTM E1926-08(2008). The quarter-car model was chosen over other models (i.e., half-car or full-car models) to define the point-wise mechanical responses of a single profile.

Terrestrial Laser Scanning and Photogrammetry Comparison:

To compare the point clouds obtained from terrestrial laser scanning and photogrammetry, the coordinates of the clouds for the same section on the same date were matched in CloudCompare (Girardeau-Montaut 2015).

The clouds were discretized and organized following the same approach described previously. Each longitudinal strip from the TLS grid was matched with a strip from the photogrammetry grid.

 TIME-LAPSED LASER SCANS

Following the described setup and analysis, five sections were scanned and analyzed over time. Three unpaved sections were scanned over time; the streets serve residential areas and farms with a few agricultural facilities.

The scanning period began after the section was bladed, and scans were taken whenever possible. Two paved sections were scanned during construction, and scans were acquired for each layer in the construction process.

 TERRESTRIAL LASER SCANNING VERSUS PHOTOGRAMMETRY

Five point clouds were compared to evaluate TLS and photogrammetry. Two clouds (5m long) were acquired at the west station of 210th Street in Boone County, Iowa. One cloud (15m long) was acquired at the east station of 210th Street in Boone County, Iowa. Another cloud (34m long) was acquired on 520th Avenue in Story County, Iowa. The fifth cloud (46m long) was acquired on Airport Road in Ames, Iowa.

 SOURCES OF ERROR AND LIMITATIONS

Errors can be divided to two types : bias and variance. Bias is a consistent shift in measurements, and variance is the noise in the point cloud. The variance in a point cloud collected using a given device can be evaluated by scanning a flat surface many times and evaluating the spread of the points about the surface.

Bias can be tested in the same way. However, to test the bias and variance for the purposes of road roughness evaluation, the point clouds should be compared to a reliable set of ground truth data collected using a reliable setup. This ground truth testing is necessary because the acceptable level of errors depends on the application.

CONCLUSIONS AND FUTURE RESEARCH

The study described in this report presented the tools to evaluate the roughness of different road types using 3D point clouds collected using two different technologies: terrestrial laser scanning and photogrammetry.

The research indicates that the technologies described are very promising for evaluating road roughness. The major advantage of both technologies is the large amount of data collected, which allows the evaluation of the full road surface. Based on the analysis conducted, the following conclusions can be drawn:

  • 3D laser scanning and photogrammetry techniques are powerful tools to provide detailed measurements, and with proper calibration they can be used in the analysis and characterization of pavement roughness.
  • The analysis was automated, and the analysis technique can be developed further to be fully autonomous starting from the scanning stage.
  • The proposed analysis technique can be used to identify localized rough features.
  • Based on analysis of the results, IRI for relatively short sections is highly variable across the width of the section.This finding warrants further research.
  • The proposed technique warrants additional development and verification to be standardized and used as a valuable tool for evaluating the rideability and smoothness of road surfaces.

Additional research should also include studying technologies for the automation of image collection. Images taken by unmanned aerial vehicles (UAV) for use in photogrammetry are already being widely used in agriculture, ecology, and mining applications.

Applications in civil infrastructure inspection are already being explored for pavement crack detection and bridge beam deformation (Ellenberg et al. 2014). Vehicle-based camera systems for road monitoring are also potentially very useful in road inspection(Kertesz et al. 2007).

Source: Iowa State University
Authors: Ahmad Abdulraheem Alhasan | Kyle Younkin | David J. White

Download Project

>> Latest Vehicle Automation Projects for Engineering Students

Download Project:

Fields with * are mandatory