14.0 CONSIDERATIONS FOR COMMON MOBILE LIDAR APPLICATIONS
Print Friendly, PDF & Email

This section serves as a brief summary of considerations for using mobile LIDAR systems (MLS) for common transportation applications. For more information about these and other applications, please see the Literature Review in Appendix A. Note that in some cases, MLS data would need to be supplemented with additional data from static or airborne scanning, or conventional terrestrial surveys for optimal results.

14.1    Generic considerations

  • While the MLS will capture objects within range and line of sight of the vehicle, non-visible objects will not be mapped. For example, the bottom of drainage ditches may be difficult to see in the MLS dataset.
  • Consider implementing a rolling “slowdown” to minimize vehicles blocking the scanner view.  This will improve data completeness and reduce artifacts in the point cloud.
  • Highly reflective surfaces at close range can sometimes be problematic, creating saturation and blooming effects (See discussion in Appendix A).
  • Dark surfaces at long ranges are problematic for some scanners because they do not reflect light well.
  • Wet pavements will generally yield poor scanning results, as do conditions where refraction is present, for example due to steam, precipitation, or heat rising from surfaces.
  • MLS do not penetrate water.
  • One can “see” noise in a point cloud because of the high resolution; however, there is also be “noise” in other survey devices such as total station data that you do not see because the points are spaced very far apart (several to 10’s of meters).
  • Many algorithms for data processing are in research and development. Hence, much processing is currently either semi-automatic or manual, depending on the application. Few completely automated procedures exist and those that do are in often specialized software packages.  However, automated ground and other surface extraction algorithms generally work well.
  • Scanning geometry (position and orientation of scanner with respect to object of interest) determines how well objects are captured.  For example, specialized systems exist to capture very detailed pavement surface data, but are not configured to acquire data on surrounding features.
  • The points in a point cloud natively do not have attributes other than XYZ coordinates and intensity values. RGB color from co-acquired imagery can be mapped to the point cloud through automatic processes.  However, attributes such as what the point represents (i.e., point classification) are applied later through manual, semi-automatic, and/or automatic processing.
  • The identification of features is usually done through virtual surveying (i.e., point selection in computer software) in semi-automatic and manual processes. Many algorithms are currently in research and development.
  • Co-acquired imagery mapped to the point cloud can be a valuable tool for measurements.
  • MLS data will be limited to a narrow window (typically <50-100m) surrounding navigable roads or relatively smooth terrain. Hence, a terrain model may need to be supplemented by additional data sources such as airborne LIDAR.
  • Not all points will have the same accuracy and the point cloud will have noise in it.  Hence, it is good practice to avoid using isolated points when making measurements.  A key strength in LIDAR data is the relatively high point density.
  • Requires the highest accuracy and point density.
  • Enables a baseline dataset for comparison of 3D design alternatives.
  • Provides detailed documentation of as-built conditions.
  • Potential 3D data source for Automated Machine Guidance (AMG)
  • Substantial time is required to develop highly accurate, detailed models from point cloud data. Rough, generalized models can be obtained relatively easy.
  • Some semi-automatic processes exist, but much of the processing is manual. Objects with simple, standardized geometric shapes (e.g., planes, spheres, and cylinders) are easiest to extract.
  • MLS provides an abundance of sample points, which individually may be of lower accuracy compared to points acquired through conventional surveying (e.g., a total station); however, collectively, they can often more accurately model an object due to the dense sampling, which allows the capturing of more detail.
  • Models simplify and reduce the data and can improve local accuracy by filtering out noise. However, CAD models are based on simple geometric primitives (e.g., line, curve, plane, cylinder, sphere, etc.). Hence, extracted features will be forced to fit a predefined shape. In the real world, objects have deflections, bulging, and other effects that can be lost when converted to a simplified model.
  • MLS data provides critical geometric information and spatial relationships to aid in planning decisions.
  • MLS data can be virtually explored by planners to reduce the need for site visits.
  • MLS data enables both qualitative and quantitative analyses of roadway quality, particularly when combined with imagery.
  • Intensity measurements are very helpful in distinguishing between damaged (e.g., cracks, spalling, staining) and undamaged sections of concrete.
  • See discussion of General Mapping/General Measurements and Modeling
  • MLS is one of the fastest techniques to acquire data for a DTM of a road and surrounding area.
  • Point cloud data is often subsampled or statistically filtered to create a DTM that will perform well in CAD or other engineering packages, which may not be designed for large datasets (e.g., file size, number of vertices).
  • Breaklines will need to be extracted semi-automatically or manually, if desired.
  • In many cases, TINS will actually be 2.5D datasets, not 3D. Hence, they will not model details on vertical surfaces (e.g., building, steep slope face) in the point cloud.
  • While CAD and GIS software offers support for point clouds and high resolution TIN models, many engineering analysis and design packages may not support the high density TINs created by LIDAR.  A few potential solutions are:

14.1.1 General mapping/general measurements (2B)

  • The identification of features is usually done through virtual surveying (i.e., point selection in computer software) in semi-automatic and manual processes. Many algorithms are currently in research and development.
  • Co-acquired imagery mapped to the point cloud can be a valuable tool for measurements.
  • MLS data will be limited to a narrow window (typically <50-100m) surrounding navigable roads or relatively smooth terrain. Hence, a terrain model may need to be supplemented by additional data sources such as airborne LIDAR.
  • Not all points will have the same accuracy and the point cloud will have noise in it.  Hence, it is good practice to avoid using isolated points when making measurements.  A key strength in LIDAR data is the relatively high point density.

14.1.2 Engineering surveys (generic discussion) (1A)

  • Requires the highest accuracy and point density.
  • Enables a baseline dataset for comparison of 3D design alternatives.
  • Provides detailed documentation of as-built conditions.
  • Potential 3D data source for Automated Machine Guidance (AMG)

14.1.3 Modeling (1A)

  • Substantial time is required to develop highly accurate, detailed models from point cloud data. Rough, generalized models can be obtained relatively easy.
  • Some semi-automatic processes exist, but much of the processing is manual. Objects with simple, standardized geometric shapes (e.g., planes, spheres, and cylinders) are easiest to extract.
  • MLS provides an abundance of sample points, which individually may be of lower accuracy compared to points acquired through conventional surveying (e.g., a total station); however, collectively, they can often more accurately model an object due to the dense sampling, which allows the capturing of more detail.
  • Models simplify and reduce the data and can improve local accuracy by filtering out noise. However, CAD models are based on simple geometric primitives (e.g., line, curve, plane, cylinder, sphere, etc.). Hence, extracted features will be forced to fit a predefined shape. In the real world, objects have deflections, bulging, and other effects that can be lost when converted to a simplified model.

Next Section >> 14.2: Project Planning