Facilities rectification to GPS survey - At warp speed
Douglas Novy Sanborn 1935 Jamboree Dr. Colorado Springs, CO 80920 Email: dnovy@sanborn.com Most municipalities have existing digital maps of utilities and parcel data. What happens when new ortho photography is acquired and the existing digital data is not spatially correct when overlaid with these orthophotos? Denver Water faced this question. Denver Water, not unlike most utility departments throughout the country, developed its GIS with the goal of building a digital replication of existing paper sources for more efficient map updates and data distribution. The City acquired orthophotography and GPS data, however, the existing GIS simply didn’t meet positional accuracy standards. Rebuild or adjust existing data was the question. Denver Water was interested in determining the exact locations for their visible facilities and repositioning their entire digital dataset based on these new GPS locations. Denver Water and Sanborn combined forces to design a method of warping the existing utility features to their exact GPS locations without affecting the network’s connectivity and modeling capabilities. This method proved so successful, that the Facility adjustment was followed up with the adjustment of Denver Water’s parcel data and sixteen other coverages used as background for facility data. Ultimately, this project has shown how careful implementation of warp transformations can maximize efficiency and minimize manual cleanup. History Denver Water started hand mapping on linens in the early 1900s. Conversion to digital mapping started in 1985 and completed in 1989 on the Informap system. This mapping system was built with the goal of making a digital replication of existing paper sources for more efficient map updates and data distribution. Denver Water stepped into the GIS world in 1996 with the migration to ESRI’s ArcInfo librarian. The data could be accessed via Denver Water’s Intranet for operations support, but engineering projects could not use the GIS as a starting point because of the spatial inaccuracy. Orthophotography showed Denver Water’s data was not positionally accurate. Engineering needed positionally accurate data for design work, and they were finding many of the mapped features were more than 80 feet from their true location. A further need to update was due to the GIS data being incomplete. Over the years updates and changes had been made to the actual water pipeline network, but not all updates had been mapped in their GIS. Pilot The decision was made to update and rectify the data, but there was some debate as to what would be the best method for Denver Water’s needs. A decision was made to attempt to warp the entire network to the features’ true GPS location. Both Real-Time Kinematic (RTK/GPS) and Differential Global Positioning System (DGPS) were tested during the pilot to see which would be the preferred capture method, and various warps were performed to determine the best method for rectification of the data. The pilot project proved that the data could be adjusted satisfactorily. RTK/GPS was the chosen method because of its higher accuracy and better reliability. The pilot also showed the facilities would best be rectified using a rubbersheet warp. Denver Water decided a rectified network would be awkward if the reference cadastral data did not fit. Denver Water asked if the cadastral data could then be fit to the facilities data, so Sanborn came up with a way to use the parameters of the facilities rubbersheet warp to warp the reference cadastral data. Work plan project specifics Step one was for Denver Water to first clean and update their data as well as possible before delivery to Sanborn. This would minimize potential problems from trying to adjust an incomplete network. Updating the data with network changes Denver Water already had would also minimize the number of new features found by Sanborn’s survey crew. Denver Water then divided the city into 16 areas with 62,000 features over 179 square miles and posted data to an Internet project collaboration web site for download. Sanborn downloaded Denver Water’s data and imported it to Sanborn’s proprietary software format (APS). This software was chosen because APS Has the Capacity to:
A collaborative effort between Denver Water employees and Sanborn crews was used to capture the GPS locations. The Denver Water employees were particularly helpful in identifying features in areas where the existing data had greatest spatial inaccuracy, greatly reducing the number of features not found by Sanborn’s survey. New features, when found, were collected and delivered to Denver Water for future inventory. Upon completion of an area, the GPS survey crews would send the point coordinates to Sanborn & Denver Water, via the project collaboration web site. Sanborn imported the collected points to APS. Sanborn’s next step was to warp only the facility data without compromising the cadastral data. To accomplish this, the water network, associated text, and one set of warp seed points were separated into a separate file from cadastral data. A Rubbersheet warp transformation was used to adjust the water network to the new GPS locations. The rubber sheet warp transforms three or more source points to the same number of control points with zero residual errors. The transformation is piecewise linear: the algorithm forms a triangular irregular network (TIN) from the source points and maps points in a triangle using a local affine transformation defined by the source/control point pairs that make up the three triangle vertices. Note that the transformation is continuous inside the TIN, but is non-continuous outside the TIN (transformations that fall outside the TIN are extrapolations). In other words, the TIN should contain the boundary of the point set being transformed. Basically a rubbersheet is best visualized as the name implies. The water network is treated as if drawn on a sheet of rubber. By moving each feature to sit atop its true location, the water pipes and neighboring features are stretched and skewed based on the movement of surrounding features, but the network is not be broken. The network was warped using only water valve points. Butterfly and conduit valves are offset from their surface features, so their true locations could not reliably be captured from the surface. Although captured, Fire hydrants were also omitted from the water network warp. Hydrants had been digitized into Denver Water’s data based on a different set of requirements and were found to be much farther off of their true location than were water valves. The fear was that features neighboring hydrants would be overly skewed if hydrants were used as warp locations. Once the network was warped, the hydrants were programmatically moved to their true locations by reassigning their xyz coordinates to the true GPS coordinates. Their pipes were then “snapped” to reconnect with the hydrants. Hydrants were next visited to reconnect hydrant branches perpendicular to the main, per Denver Water’s requirements. Sanborn’s next step was the warp of cadastral data based on average warp of the water network. The cadastral data had been previously separated, with cadastral annotation and one set of warp seed points, into its own file. The warp seed points in this file were warped to the points moved during the water network’s rubbersheet warp. This cadastral warp was done using a Helmert 2d transformation. Using this warp, the cadastral data was moved based on the overall average facilities warp for a given area. The Helmert_2d transformation is one of the most conservative warp transformation algorithms. It uses a single variable scale factor (homogeneous shrinkage or expansion), rotation, and translation, so the integrity of the parcels would not be compromised. Although the final warped locations would not be as exacting as those of the rubbersheet warp, square parcels would remain square and blocks of parcels would be moved near where they needed to be. It was assumed there would be some manual cleanup to the cadastral data. Warp seeds, both the original locations and the locations after the network warp were saved in a separate location saved for use later in warping project-wide coverages. A manual review and cleanup of entire dataset was performed after this second warp. Denver Water’s criteria included:
Once all potential problem areas were reviewed, the data was converted back to ArcInfo coverage format, with the new features Sanborn’s survey found separated from existing features through attribution. Data was then delivered to Denver Water’s QC contractor and Denver Water via the Internet project management site. 30 QC steps were completed including surveying 3 random checkpoints per section to check for accuracy of collected points. Upon passing the 30 QC steps, acceptance or rejection was recommended (no areas have been rejected to date). Issue & resolution For a project as large as this, there were relatively few problems. Prior to surveying, permissions were needed to set up the control network. Not a problem per-se, but redtape slowed down the process until all city agencies issued permits. One problem that was recognized early-on was the Downtown area was not conducive to RTK/GPS survey due to tall buildings blocking the satellites. It was agreed that the downtown area was defined for use of Total Station collection (conventional survey method). Some features could not be found due to new pavement, overgrowth of landscaping etc. Denver Water employees used Schonstedts, which are metal detector like instruments to locate these obscured features. Features under cars were revisited on another occasion. Features under trees or landscaping were collected using Total Station. Areas crowded with features caused problems with incorrect linking of coordinates to database feature. Because some of Denver Water’s features were located up to 80 feet from their true location, the survey crews had some difficulty determining exactly which feature on the map related to the feature they were about to GPS. Denver Water employees collaborated with Sanborn’s survey crews with assistance in the field, and education of how features are laid out. On occasion, one of Sanborn’s technicians would find a point more than 100 feet from the GPS location. This happened when the data was originally digitized and a feature was placed in an area without any landmarks for the digitizer to use in determining distance. Warping one point over 100 feet could abnormally skew the surrounding data. It was agreed that any feature over 100’ was left out of the adjustment process and manually re-visited to determine what movement was needed. All large adjustments were reviewed with Denver Water to determine resolution through the PAR process Results The results have been a big success for Denver Water. Denver Water is now confident that the digitally stored locations of their surface-visible features are accurate to within 3cm of their true locations and can be the starting point for engineering design work. Also, the entire dataset has been repositioned to its true location, so most users wont notice the transformations to the GPS locations even took place. And now the database is more complete because new features found in the field have been added. This project has shown how careful implementation of warp transformations can maximize efficiency and minimize manual cleanup. The network of water features was warped to the features’ true locations without affecting connectivity, and cadastral features were warped without affecting their integrity. Warping also kept the manual efforts to a minimum, providing a cost-effective solution to Denver Water’s original dilemma of spatial inaccuracy. | ||
|
|