TOPOGRAFIA y CARRETERAS

Terrestrial 3D Laser Scanners

Purpose

This chapter provides an overview of 3D laser scanners used for detailed mapping of facilities, structures, utilities, and ground planes. High-precision/high-definition tripod mounted laser scanners are covered. Examples of recent applications where these instruments have been used to map Corps facilities or structures are provided.

Figure 10-1. Optech ILRIS 3D laser scanner--detailed mapping of structures and facilities at St. Lucie Lock and Dam, Jacksonville District (Arc Surveying & Mapping, Inc.)

Background

Laser scanners operate similarly to reflectorless total stations. However, instead of a single shot point being observed, a full field-of-view scan is performed--at a speed upwards of 500,000 points per second. Unlike a total station, the location of the scanner is not usually a required input--resulting in points that are spatially referenced to the instrument and not real-world coordinates. This is somewhat analogous to an uncontrolled photogrammetric model. (Newer models allow input of the scanner coordinates from which all observed pixels may be directly georeferenced). The resultant imagery from a scan (termed a "point cloud") provides a full 3D model of the facility, utility, or terrain that was scanned. Objects can be scanned at a high density--with output pixels smaller than 5 mm. Relative 3D accuracies approaching the millimeter level are claimed, based on redundant observations over a surface. However, 5 to 10 mm accuracy is more realistic in practice. Scans can be rapidly made--a full field-of-view scan of a site or structure can be performed in 5 to 15 minutes per setup (multiple setups generally are required to fully detail a given site or structure). Unlike a total station, however, laser scanners have no means of assigning feature codes or attributes to the measured points--this must be done in post-processing, and is often a tedious and time-consuming process. Laser scanners have increasing application to many Corps civil and military missions. They can be used to perform traditional topographic surveys (detailed planimetry and elevations) of project sites and facilities--providing ground elevations at a high density. These scanners are especially useful in detailed mapping of exposed (and hard to access) utility systems, such as those inside a hydroelectric power plant. They also have application in mapping archeological sites, HTRW sites, dams, rock faces, hazardous traffic areas, unexploded ordinance sites, or any other inaccessible location.

Figure 10-2. Laser scanned image (upper left) and rendered image (lower right) of Corps Mississippi River Division headquarters building, Vicksburg, MS (ARC Surveying & Mapping, Inc.)

a. Manufacturers.
Laser scanners have been on the market for only a few years--since the late 1990s. As of 2005, there were about 12 manufacturers of laser scanners listed in trade publications.
b. Cost.
A complete laser scanning system (including a high-end modeling software package and training) can cost between $150,000 and $200,000. Thus, few, if any Corps Districts would have a sufficient number of applications to justify this level of expense. In time, it is possible these prices will decrease to a level where 3D scanners may be cost-effective if the workload warrants. Data processing and modeling software is typically expensive; however, it is essential in order to export scanned images to CADD platforms. At present, Corps Districts contract for periodic 3D scanning services--many AE surveying contractors performing Corps work have acquired (or are acquiring) 3D scanners. Given the limited amount of work, the hourly/daily rate is understandably high for these scanners (and data processing)--daily operating costs of $2,000 to $5,000 or more (including processing) are not uncommon, and can vary widely depending on the amount of processing required. Most often, a project cost would be negotiated on a lump sum basis, factoring in the basic daily rental cost of the scanner plus the operator and CADD processor time estimates.
c. Accuracy.
The accuracy of a scanned object can be relative or absolute. Relative accuracies are very good (5 mm or better at close ranges). Absolute accuracies depend on the accuracy of the control network developed for the site, how accurately the instrument is aligned to this network, and how well overlapping images (i.e., picture points or targets) are transferred and adjusted (best fit). In general, absolute accuracies can be kept within 1 or 2 centimeters over a small project/structure site. In many cases, relative accuracies are far more important than absolute geospatial accuracies. For example, measurement of a crack in a wall requires a high relative accuracy; however, the absolute geospatial coordinates of the crack are not significant. Overall accuracy is a function of range, scan density, spot/footprint size, and single point accuracy (Jacobs 2004).

d. Density of scanned points. Laser scanners can be set to any desired scan density, e.g., 5 mm to 1 meter, usually based on some short nominal distance. The higher the density the larger the resulting dataset--and more time-consuming data editing and processing. The purpose of the project determines the required density. For general 3D planimetry or buildings or ground elevations, a low density can be set. For detailed maps of structural members or concrete cracks, a high density is set.
e. Field-of-View.
Depending on the model and project requirements, scanners can be set to scan a full 360 deg field or zoomed (windowed) into a narrowly set field-of-view. The Leica HDS4500 scanner has a field-of-view of 360 deg horizontal by 310 deg vertical. The reality of scanning means that even a 360º field of view does not guarantee full area coverage; in fact, it rarely does. Laser shadowing forces data to be collected from multiple angles for complete data coverage. A large field of view does not alleviate this requirement. Also, the angle of incidence of a measurement will have a profound impact on its accuracy and the resolution of the data in general. Keeping this fact in mind, imagine surveying a long, flat wall. If the scanner has a 180º horizontal field of view it will be able to survey the wall in a single scan. Depending on the length of the wall, measurements may be collected from angles that approach 90º, but the reliability of these measurements will be very poor.
f. Range.
Range of scans is a function of the laser intensity and reflectivity of the object scanned. Some scanners are designed for only close up scans--i.e. 200 meters. Others claim ranges of upwards of 1,000 meters--or more. Obviously, the longer the range, the larger the footprint and less accurate the resultant measurement becomes. In general, most detailed scans of facilities, buildings, and structures are kept at close range--usually less than 500 ft and not much beyond 1,000 ft. The laser eye safety classification may also be a factor in longer-range scanners--a Class 3 type laser may not be desirable for surveying a populated beach but would be acceptable at a remote HTRW site. (A Class 1 laser device "denotes exempt lasers or laser systems that cannot produce a hazard under normal operating conditions" and a Class 3a laser device "denotes visible lasers or laser systems that normally would not produce a hazard if viewed for only momentary periods with the unaided eye. They may present a hazard if viewed using collecting optics."). The ranging capability of a laser scanner is more important than it first seems. In some cases, the lack of ranging ability will completely eliminate the ability to do certain projects. For example, let’s examine the case of a bridge that crosses a body of water. A lack of certain characteristics (such as a 360º field of view) may force you to collect more scans but it doesn’t prevent the project from being completed. The inability to range to the structure, with no provisional means of getting closer, will eliminate the potential of the scanner being used on the project. It is also true that a survey will rarely be conducted at a range greater then a few hundred meters (except in cases similar to the above example). However, long-ranging capability has other benefits that are not immediately identifiable. At the extreme limit of a scanners specified range, the accuracy of the measurements will begin to decrease in a non-linear fashion. As such, let’s consider the example of scanning a structure from 150 m away. The scanner whose maximum range is 150-200 m will struggle to collect the data. It may succeed, but the measurement quality will suffer and the dataset will contain a large number of "drop out" points (instances where no measurement was collected). Alternatively, the scanner that provides 1,000 m range will be collecting data from the ideal area of its total dynamic range. The data collected will be of optimum accuracy.
g. Beam footprint size.
The footprint size will vary (increase) with the distance from the scanner to the object. Typically, on close-range applications (less than 100 ft ranges), a 5 mm footprint is observed.
10-3. Scanner Operation and Data Processing

Scanners are normally mounted on a tripod, directly onto the plate or in a standard tribrach. The scanner is set up at any arbitrary location that affords the best view of the area or object to be mapped. No absolute geospatial orientation of the scanner is required (unless the scanner model is designed to incorporate geospatial references). Most structures require multiple scans in order to develop a complete 3D model, as illustrated in Figure 10-4 below. In addition, multiple scans are required to cover hidden, shadowed, or obstructed areas in a single scan. Thus, a rectangular building will require scans from four setups offset from each of the corners--each scan providing data covering two faces of the building, which will overlap with adjacent scan locations. These overlapping scans allow a full 3D model of the building to be generated using imagery correlation (optical recognition) software (similar to soft-copy photogrammetry). The scanned "point clouds" are saved on flash memory devices in the scanner, which can be later downloaded to a field or office computer. The overlapping "point clouds" from each scan are edited for data spikes--often a lengthy process. They are then merged to form the full 3D model. This merging is done in proprietary software that is usually sold separately with the laser scanner. This resultant 3D model is referenced only to a relative/internal coordinate system. If real-world geographic X-Y-Z coordinates are required (and they are not always needed for many project applications), then targeted points need to be set in the scanned area/structure in order to perform a standard coordinate transformation. Once the model is generated, a variety of computer graphic enhancements can be performed. These include coloring, wire meshing, rendering, and smoothing objects. Rough point cloud images of solid objects can be smoothed using various software-fitting routines--e.g., items such as wall
faces, cylindrical pipes, etc. If the resultant model is going to be exported to a CADD or GIS platform, then additional descriptor, attribute, or layer/level assignments may be required. Final data processing can represent a significant effort on some projects--a structure that is scanned in 4 hours may take as much as 40 hours or longer to process the data to a CADD compatible format. The software used for processing scanned datasets is a critical component in the overall efficiency and economy of the process.