Double the Power
July 1, 2006
LiDAR, the technology that determines distance to an object or surface using laser pulses while collecting mass point clouds, has gained in popularity over the years. Clients, however, continue to have concerns about the technique and ask questions about this powerful technology. At Vertical Mapping Resources (VMR) Inc., our clients regularly ask us what applications it can be used for and what sort of accuracy can be achieved with it. They want to know how to handle the copious amounts of data derived from LiDAR processes and how to utilize the technology on future planned projects.
These concerns are understandable, as LiDAR is still a relatively new technology and vastly different from other techniques. In large-scale mapping, photogrammetrists generally fly at about 1,800 ft above the ground. In the mapping process almost every visible feature is collected, including roads, buildings, manholes, utilities, valves, poles, etc. Surface modeling and contours are also provided, generally meeting a .5 ft vertical accuracy requirement. This is in addition to digital othophotography, with nominally a .3 in pixel resolution. It's a long process that involves many steps. But with the advent of LiDAR, mass point clouds are collected containing everything the laser hits.
Since most of our projects deal with large-scale mapping, it's hard to find a way to utilize LiDAR data that is cost-efficient because many of our projects also include planimetrics and digital imagery in addition to surface modeling. At VMR, we have developed a solution to handle these job parameters by combining LiDAR with photogrammetry. Our solution results in highly accurate data sets.
The best fit we have developed for LiDAR together with photogrammetry is on a project that requires both large- and small-scale mapping. A client might have an area of interest (AOI) to be mapped, but the surrounding terrain plays a large role in his or her decision of which technology to use. For example, say a client requires design scale mapping with 1 ft contours for several hundred acres to design a new housing development with transportation corridors. The client needs to evaluate the surrounding area to determine flood patterns and right of way, but does not require design scale mapping for these outlying areas. Since budgets are almost always a concern for clients, 2 ft or 5 ft contours would easily supply this client with the information required.
Over the past year, we have performed several projects identical to this scope. Our methodology is to fly LiDAR, collecting mass points that cover the entire area. This provides cost-effective surface modeling in a digital elevation model (DEM) format that the client can use for planning purposes in an early stage of the project, without taking a large bite out of the project budget. Next, we perform aerial photography and photogrammetric mapping for the AOI. Matching into the LiDAR DEM, we supplement it with breaklines, collect planimetrics and provide ground-stitched orthophotography. A large benefit of flying the LiDAR first is that it allows us to utilize the mass point data in the design scale mapping. Without this technique, mass point data would need to be collected manually, which is quite time-consuming. Instead of a proposal to capture data one way or the other-LiDAR versus softcopy photogrammetry-we propose a combination of the two technologies. Here are the specific steps taken to combine these two great technologies.
LiDAR CollectionTo begin a mapping project, we will commission our LiDAR partner to collect the mass point DEM as soon as the project ground control has been placed. We design one layout for sufficient ground control for both the LiDAR and aerial photography flights.
The major components we utilize for laser collection include: a Partenavia P68 aircraft; an Applanix (Richmond Hill, Ontario, Canada) POS/AV airborne vehicle with POS/Pac AIR post-processing software with GPS/Inertial Measurement Unit (IMU); and an Optech (Toronto, Ontario, Canada) ALTM2050 LiDAR system. The overall system consists of a geodetic GPS positioning unit with orientations derived from high-end IMUs and a powerful laser. The LiDAR technology offers fast, real-time collection of three-dimensional points that are employed in the generation of a DEM.
Once the flight is completed and the LiDAR information has been collected, extraction of bare earth topography from the raw LiDAR data is accomplished by running an iterative morphological classification algorithm. This algorithm examines the data points and classifies them as "ground" or "extracted features" based on distance and angle of each point from an initial Triangulated Irregular Network (TIN) surface. The reliability of this modeling may vary with terrain relief, vegetation cover, cultural features and the setting of initial seed parameters for the classification algorithm.
The end result of the LiDAR data collection process is millions of points defined by their earth-centered, earth-fixed Cartesian coordinates. These points are converted to grid coordinates and elevations. This process involves the integration of the ground control, laser information and the IMU with spatial models. The points classified as ground are separated from the rest of the clutter data and considered as the LiDAR DEM, or "bare earth" LiDAR data.
Since the LiDAR point data is so dense-oftentimes gigabytes of point data while still in ASCII format-the files are much too large to easily manipulate in a CAD environment. To significantly reduce these large file sizes, the LiDAR DEM is converted into an interpolated fixed grid with 10-ft ground point spacing for integration into the photogrammetric mapping portions of the project.
LiDAR data is normally checked using blind panels and cross-sections for ground checks to determine an overall Root Mean Square Error (RMSE). In addition, since we utilize the data for our photogrammetric work, all LiDAR points in the interpolated DEM will be visited stereoscopically on our softcopy photogrammetric software to see if each individual point is, in fact, on the ground.
Photogrammetric CollectionSince we utilize two different aircraft on these types of projects, we fly our photogrammetric aerial mission right after the LiDAR flight is complete. This way, not only are we guaranteed that our ground control panels are intact, but we can get started on our internal production schedule. This, in turn, allows our compilation department to be ready when the LiDAR DEM is completed and QA/QC has been performed. For ease of use, we design complex aerial projects like this so adjacent flightlines run in a cardinal direction, with each flight line oriented in the same forward direction. This makes things much easier when comparing a large block of LiDAR data to a large block of aerial photography, especially when the photography consists of several hundred stereomodels.
The major components for our photogrammetric mapping include: our twin-engine Cessna 310Q aircraft; an Applanix POSAV GPS/IMU; a Carl Zeiss RMK Top15 camera; an Intergraph (Madison, Ala.) PhotoScan scanner with roll feed; Intergraph Softcopy Software, including ImageStation Digital Mensuration (ISDM), ImageStation Stereo Display (ISSD), ImageStation Feature Collection (ISFC), ImageStation DTM Collection (ISDC) and ImageStation OrthoPro; and a Bentley (Exton, Pa.) MicroStation v8 platform.
Once we receive the film from our aerial partner, it is scanned on our PhotoScan directly from the roll. We run all of our film through our film cleaner to be certain the majority of foreign particles have been removed. Our scan resolution is normally 14 microns, but from past experience we scan projects of this scope at 7 microns to ensure that our technicians can see the ground at the best possible resolution when checking the LiDAR DEM.
Next, we perform analytical aerotriangulation utilizing a combination of Intergraph's ISDM/ISAT modules. This combination, in our opinion, is one of the most robust triangulation packages on the market. It has enhanced tools for point measuring, both manual and automatic. Our bundle adjustments usually have around 100 pass points per exposure and anywhere from 10 to 75 tiepoints based on the terrain. We use a 5x3 Von Gruber pass point schematic on all of our projects; the diagram on page 26 shows the layout for optimal pass point selection over a photograph and its adjacent photographs in a stereomodel. As a solid check, our adjustments are compared to the project ground control on every project. Each panel location is visited; bundle adjustments don't pass until every control point sits dead-on to its true location stereoscopically.
Finally, we are ready for softcopy compilation and merger with the LiDAR DEM. For the purpose of this article, we will focus only on the DTM creation, where we take the DEM from the LiDAR, check it and then supplement with hard and soft breaklines.
Why are we supplementing the DEM? For design/large-scale mapping, the contour interval is normally 1 ft. To meet the ASPRS Class 1 or Federal Geographic Data Committee (FGDC) requirements for vertical accuracy, breaklines are required for all surface modeling and contouring. The 10 ft fixed-grid DEM will not meet these requirements on its own. For a graphical depiction of this explanation, look to Figure 1a, which shows a shaded relief of the interpolated LiDAR DEM after it has been triangulated into a TIN. Figure 1b shows a shaded relief of the interpolated LiDAR DEM supplemented with manually collected breaklines, or the DTM after it has been triangulated into a TIN. As seen in the figure, the LiDAR DEM gives a resemblance of the roads and hard cuts in the ground, but the breaklines from the DTM are required to accurately model the terrain and meet mapping standards.
There are three major steps to arrive at the completed DTM. First, the LiDAR DEM must be checked stereoscopically against the stereo imagery to ensure that all points are accurate and on the ground. Next, breaklines and formlines are added to the DEM per VMR's normal design-scale collection standards. Finally, for our QA/QC, 1 ft contours are generated and everything is again checked stereoscopically and edited as necessary until both surface and contour data are on the ground. With softcopy photogrammetry, QA/QC is always constant thanks to the superimposition and stereoscopic environment.
In Figure 2, the LiDAR DEM is shown superimposed over the stereo imagery. In softcopy photogrammetry, superimposition is the process of overlaying vector data within the stereomodel. The vector data retains its coordinate information (X,Y,Z) and sits in the model space along with the stereo imagery within the project coordinate system. Points on the ground will be seen as such. However, points above or below the ground will "float" or "dig" on the imagery and are easy to identify. These points are, of course, off the ground and deleted from the DEM.
Figure 3 shows the completed DTM superimposed over the stereo imagery. The DEM has been supplemented with breaklines collected along the road edges, drainages and pads. Formlines have been collected along the slopes and flatter areas. The breakline and formline information is what supplies the TIN with the data it needs to accurately contour at a 1-ft interval.
Figure 4 shows two sets of generated contours. Since the terrain is very steep, we have only displayed the index contours so differences between the two data sets can easily be seen. The blue contours have been generated from the LiDAR DEM only. This data set contained no breakline information. The blue contours cross the roads but do not accurately model the surface of the pavement. The red contours have been generated from the DTM created, which contains both LiDAR and breakline information. This surface has accurately modeled the road, and the red contours trace both sides of the roadway and flatten across the pavement of the actual road. Closer analysis will also detail discrepancies in the hilly areas between both data sets.
Figure 5 shows another example of the differences between the two data sets, this time shown in a 3D perspective view. Again, the blue contours are from the LiDAR DEM and the red contours are from the combined DTM that consists of both LiDAR and breaklines. The 3D view shows that lacking surface information will lead to not only vertical error but horizontal error as well.
Project ConclusionsThe result of using these two revolutionary technologies-LiDAR and photogrammetry-in concert is a very accurate photogrammetric data set. Instead of using one technology or the other, it is a marriage of the two that focuses on the strengths of each technology. The end result is an ideal combination of small- and large-scale mapping, created with benefits that lower both cost and production schedules for the client. The use of the LiDAR provides an accurate, low-cost surface model for planning, land development and GIS departments assigned to these types of projects. And the supplementation of the breaklines, planimetrics and final orthophotography provide the civil and transportation engineers with the highly accurate data they require for design purposes.
In addition to achieving accuracy and mapping standards that were above the requirements for projects VMR completed using the above method, project data was delivered ahead of schedule in each case, and the clients received LiDAR contours and scratch orthophotography within two weeks of VMR receiving the LiDAR data. These projects are shining examples of the combination of industry proven photogrammetric principals, along with the cutting edge technologies of softcopy photogrammetry and LiDAR DEM collection.
Sidebar: Glossary of terms:An algorithm is a step-by-step problem-solving procedure (especially an established, recursive computational procedure) for solving a problem in a finite number of steps.
Aerotriangulation is the process of assigning ground control values to points on a block of photographs by determining the relationship between the photographs and known ground control points.
Analytical aerotriangulation is aerotriangulation in which coordinates of ground points are derived from measured coordinates of image-points using mathematical functions rather than a physical model to represent the relationship between image-point and object-point.
Blind panels are surveyed panel points that are not included in the aerotriangulation solution and are used for final checks as a known point.
Ground-stitched orthophotography is orthophotography that is stitched (mosaicked) together with seamlines that are manually drawn between images and follow the ground and avoid above-ground features.
Large-scale mapping is commonly mapping at a scale of 1:50000 or larger.
LiDAR (Light Detection and Ranging) is a technique that uses an instrument capable of measuring distance and direction to an object by emitting timed pulses of light in a measured direction and converting to the equivalent distance the measured interval of time between when a pulse was emitted and when its echo was received.
A micron (short for micrometer) is a unit of length defined to be exactly 0.000001 meter.
A morphological classification algorithm is an analysis technique developed by Fritz Zwicky (1966, 1969) for exploring all the possible solutions to a multi-dimensional, non-quantified problem complex.
Orthophotography is the art of preparing a
true vertical photograph of a landscape. An orthophotograph is prepared from a perspective photograph by removing those displacements of points caused by tilt, relief and central projection.
Photogrammetry is the science of deducing the physical dimensions of objects from measurements on photographs.
A planimetric map presents only the relative horizontal positions of natural or cultural features by lines and symbols. It is distinguished from a topographic map by the omission of relief in measurable form.
Root Mean Square Error (RMSE) is determined by calculating the deviations of points from their true position, summing up the measurements, and then taking the square root of the sum.
Small-scale mapping is commonly mapping at a scale smaller than 1:50000.
Stereoscopy is the art and science dealing with the principles and applications of the ability of the human visual system to create a mental image of a three-dimensional object from two separate, two-dimensional images as seen by the eyes.
A stereomodel is a three-dimensional image, either actual or visual, formed by the intersecting, homologous rays from the corresponding points in the separate images of a stereoscopic pair of images. Also called stereoscopic model and spatial model.