A Benchmark for Data Fusion
With so many technologies available for collecting geospatial data, it is increasingly difficult for contracting entities to decide which technology or data source will give them the best possible data for a given project. Understanding that each dataset offers specific benefits, many clients have begun asking for deliverables that are obtained through a combination of different technologies and datasets. This process, commonly referred to as data fusion, is generating a great deal of excitement throughout the entire geospatial industry. However, there are still many questions about data fusion and its use.
One of the biggest concerns facing the end user is accuracy. Different data collection methods yield different accuracies, depending on the technology that was used during the collection phase. Merging datasets with accuracies that vary from 1 to 10 centimeters can be cumbersome, to say the least. To fully capitalize on the benefits of data fusion and achieve a successful outcome, the project team must be capable and must fully understand the advantages and limitations of each data source.
In July 2011, the Michigan Department of Transportation (MDOT) contracted Surveying Solutions Inc. (SSI) of Standish, Mich., to conduct a full topographic survey of the I-96/US 23 highway interchange in Brighton. One of the busiest interchanges in Michigan, the I-96/US 23 junction consists of several right and left exiting and entering ramps, tightly merging lanes and a long a section of high-speed curving expressway with rolling grades. Because of the high level of traffic, the aggressive six-week timeline for project completion and the need for data across the entire 8.2 miles of the project, including areas outside of the traditional right of way, collecting the data entirely through traditional survey methods would have been nearly impossible.
The only way to address these concerns while maintaining the ability to acquire highly accurate survey data was through data fusion, using a combination of mobile LiDAR, traditional ground survey methods and existing aerial photogrammetry.
As a prequalified MDOT subcontractor for road design, SSI is involved with projects that require the highest levels of accuracy. After evaluating some of the most accurate systems in the world on a stretch of highway that was heavily surveyed with conventional methods and used as a test site, SSI had acquired a Riegl VMX-250 in April 2011. With its mobile LiDAR services, which it called MoLi (pronounced “molly”), the firm was confident that it would be able to provide detailed, highly accurate surveys quickly and safely, without disruptions to the traveling public.
The GPS/INS solution integrated into mobile mapping systems is by far the largest source of error in the data. The Riegl VMX-250 is a state-of-the art system that provides a LiDAR precision of 5 millimeters, a relative accuracy from point to point of 8 millimeters and an overall system accuracy of 2 to 5 centimeters with optimal GPS. However, to achieve the accuracies demanded for engineering design, the mobile mapping system must be able to generate data better than 2 to 5 centimeters; design specifications are often to a tolerance of 2 to 3 hundredths of foot. This tolerance far exceeds the general system specification of any off-the-shelf mobile mapping system. A firm can only achieve design-grade accuracies if the LiDAR sensor itself as a component is able to achieve design-grade tolerances. By using traditional surveying methodologies to establish a control network, SSI was able to eliminate most of the error from the system’s navigation solution and start approaching the specifications of the LiDAR sensor as the main error source, which yielded accuracies and tolerances acceptable to the MDOT project team for the Brighton project.
To ensure the success of the project, SSI and MDOT spent a large amount of time in the planning phase. MDOT had previously contracted for the project area to be flown using aerial photogrammetry, and the resulting triangulated irregular network (TIN) model would be supplied to SSI for use in the non-hard surface areas. The project team decided that SSI would use traditional survey methods and the Reigl VMX-250 mobile LiDAR system to tighten up the accuracies and triangulations on all of the hard surfaces throughout the project area. This would provide MDOT with survey-grade accurate data, and a more complete and representative surface model of the interchange.
After completing and verifying the control network survey, SSI deployed its MoLi system to scan the entire project area. Best practices established by SSI during drive procedures helped ensure overall data integrity. Contrary to popular belief, simply turning the system on and driving the project doesn’t produce usable data. Each mobile mapping firm has specific regimens that operation crews must follow to achieve optimal datasets. Particular concerns during driving are data management for processing, known areas with little or no GPS, public traffic through the project site and PDOP windows for optimal GPS conditions.
After the data was collected from the mobile LiDAR system, SSI’s LiDAR technicians verified the 3D accuracy of the point cloud data by producing a “cloud to control” comparison spreadsheet showing any vertical errors throughout the project at specific control points. These errors were corrected to meet project scope requirements and tolerances. Technicians then began extracting the information on the hard surfaces using Certainty 3D TopoDOT software. Two-dimensional line work was digitized over the point cloud data to verify horizontal positioning with the original survey. SSI technicians then “draped” the line work to the point cloud to produce the required 3D line work.
It was agreed upon by MDOT and SSI to extract vertices along the roadway at 10-foot intervals in straight sections. Intervals of 5 feet were used along the curved sections of the interchange. This process provided a much higher level of accuracy and detail than traditional survey methods within the project timeframe allotted by MDOT.
After the extraction was completed from the mobile LiDAR point cloud, the data was then imported into Caice software to produce a TIN model of the hard surfaces of the interchange. This information would then be fused with the information from the existing TIN derived from aerial photogrammetry.
Within the original MDOT provided planimetric file, all the hard-surface line work was removed and replaced with the new extracted data from the mobile LiDAR, which updated several pavement changes within the project limits from the original survey. The SSI team then used Bentley MicroStation/GEOPAK to merge together the original MDOT provided TIN file with the newly produced TIN file from the hard-surface extraction, producing a final surface file. Because the project timeline didn’t allow for vegetation removal and grass mowing, SSI had collected additional survey data through traditional methods in areas with extensive vegetation and tall grass to ensure that all project tolerances would be achieved. This data was also merged into the final TIN model. By fusing the three datasets, SSI was able to produce a final DTM in a single file format for MDOT.
Using mobile LiDAR gave the MDOT project team a more thorough and representative dataset. The ability to use multiple data sources ensured that MDOT received all of the information that was collected while also giving Michigan taxpayers the best return on their tax dollars. Additionally, having access to a more representative model will help the project team mitigate many unforeseen concerns that may occur throughout the project’s life cycle.
As MDOT considers future transportation projects, the Brighton project provides a valuable model for surveying methodology. “New tools and technologies have rapidly changed the ways highway corridor projects are being surveyed,” says Kelvin J. Wixtrom with MDOT. “Each new technology provides its own unique advantages for acquiring portions of data better, or more efficiently, than other methods. Multiple tools generating several datasets are generally required to cover the entire project efficiently, [and] data fusion is becoming necessary in many more projects.
“Proper planning early in the project that addresses not just the acquisition of the data, but also looks at the software needed, the office processes and workflow to be used, will provide better success in producing a quality data fusion product. In addition, an important consideration in merging data is having overlapping datasets that can be compared for accuracy and uniformity. Each data set should be compared to the project’s main set of control or validation values to determine its appropriate use. Choosing the right location for the merge line is critical to providing a quality surface for use in design.”
For more information about the Riegl VMX-250, visit www.rieglusa.com. Certainty 3D’s website is www.certainty3d.com, and Bentley’s website is www.bentley.com.