Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
93 A p p e n d i x e The quality assurance process was a continuous task to ensure that quality data were collected by the mobile data vendor. Random sites, control sites, and a curve analysis were used to evaluate the accuracy of the data. This appendix provides, as an example, an overview of the postprocessing of the quality assurance plan for one of the NDS sites (Pennsylvania) and for one of the data collection years (2013). Similar processes were used on all the data collected by the mobile data collec- tion vendor. From a geographic information system (GIS) perspective, all roadway features are represented with individual GIS data layers. The resulting data set is essentially a set of overlaid data layers. Figure E.1 presents a roadway segment with SHRP 2 Roadway Information Database (RID) roadway features. While some roadway features are point features, such as signs and intersections, some are longitudinal (segment) features, such as shoulders and lanes. Random site data points collected by the team are points with all available roadway data attri- butes at that location. For example, data collection at point A in the figure would provide lane (type and width), shoulder (type and width), grade, cross slope, rumble strip, and lighting data. These data would then be linked to the GIS data layers from S04B vendor data during postprocessing. Before each site visit, a field preparation procedure was followed in-house by the project team. The procedure con- sisted of preparation of maps (GIS maps for the data collec- tion device and hard copies for the team), data collection equipment, and initial routing. Initial routing helped the team broadly identify areas to cover each day and the most efficient route. After site visits, collected data were submitted to the analy- sis team, and postprocessing began. Typically, 130 to 200 data points were collected at the site visits. Through a custom- made ArcGIS toolbox and scripts (Figure E.2), random data site points were matched to RID roadway-feature data layers. One roadway feature per layer was assigned to a random point, except for rumble strips and lighting since they are linear segment and not point data. These two layers assigned two features if present on both sides of the roadway. The ArcGIS toolbox also had an ambiguity and review feature, which notified the team during the accuracy analysis to inves- tigate the roadway feature further. Ambiguity occurred when the random point was at the intersection of two lines, and the additional review occurred when the user was unsure about which feature to select. These data sets were then processed to create comparison data sets for each roadway feature. The roadway features and comparison criteria were as follows: ⢠Intersections (location, type, number of approaches, con- trol type); ⢠Signs (MUTCD code, speed limit value, image); ⢠Highway lighting (presence); ⢠Lanes (type, width); ⢠Medians (presence, type); ⢠Shoulders (type, width); ⢠Rumble strips (presence, type); ⢠Location (grade and cross-slope values); and ⢠Barriers (barrier type, start/end treatment type, post material, rub rail). Accuracy requirements for length and location of features are presented in Table E.1. For each roadway feature, charts presenting whether the vendor data were correct in attribute match (e.g., lane type) and met accuracy requirements (e.g., location of a vendor speed sign in allowable distance from the site data point) were prepared (Figure E.3 to Figure E.15). For identified problems, data collection video log files were reviewed. After this refine- ment, results were communicated to the S04B team. Described below are the steps for the roadway featuresâ quality assurance for the random and control sites. The control site charts have multiple comparisons to the mobile data, which accounts for the separate trips through the control site by different units. Quality Assurance Process
94 A Figure E.1. S04B roadway features on a road segment. Figure E.2. GIS module to link data. The first step in the process was to verify coverage of all random site points. Location data were present every 21 ft and were located on every data collection route. All points were verified to have location data, even when no grade or cross-slope data were collected. The grade and cross slope were then compared to the location data to determine the differences between the random site data collected and the mobile data. These can be seen in Figure E.3 through Figure E.5. Intersections were then checked for accuracy requirements. The intersection is represented as a point and should be located along the side of the road in âdirection 5â (i.e., the primary or cardinal direction of travel). It was first verified that all inter- sections were identified as part of the mobile data collection.
95 The intersection point from the mobile data had to be within 150 ft of the random/control site point. The final check was whether the number of approaches was correct, as well as the control type (Figure E.6). The other point features from the mobile data collection were signs. The signs were separated into speed signs and all other signs, because each possesses different attributes. The presence of all signs within the mobile data were verified, as well as whether the signs were within 7 ft of the random/ control site point. Speed signs were evaluated for the accuracy of the speed limit text and the classification of regulatory and advisory sign types (Figure E.7). The MUTCD code and sign text of the remaining signs were evaluated for accuracy (Figure E.8). As discussed in Chapter 6, when running the script to join the reference field data collected and the mobile data, only one roadway feature from each layer was associated to a random/ control point. The exceptions were highway lighting and rum- ble strip data. If present, these features had segments selected on both sides of the roadway for comparison. Highway lighting Table E.1. S04B Data Accuracy Requirements Data Element Accuracy Requirement Curvature radius 100 ft (curves less than 1,500 ft radius) 250 ft (curves between 1,500 ft and 6,000 ft radius) Within 13% (curves over 6,000 ft radius) Curvature length 100 ft (curves less than 1,500 ft radius) 200 ft (curves above 1,500 ft radius) PC 50 ft PT 50 ft Grade (+ or -) 1.0% Cross slope/superelevation 1.0% Lane width 1 ft Paved shoulder width 1 ft Inventory features (signs) location 7 ft was represented on both sides of the roadway, if present, and the indication was confirmed (Figure E.9). Rumble strips on both sides of the lane were classified as a set. The presence of rumble strips was also confirmed (Figure E.10). The linear features, medians, guardrail, and lanes were all checked individually to verify correct classification (Figure E.11 through Figure E.13). Medians were first evaluated for pres- ence. Of the medians that were present, the type of median was evaluated for accuracy. Guardrails were also evaluated for the accuracy of the start and end treatments, the type of guardrail, whether a rub rail was present, and the post material. The lanes data were first evaluated for the accuracy of the width. The widths were assessed based on the required 1 ft accuracy. The number of lanes was then compared, as well as the classifica- tion of the lane types. Shoulders were separated into paved and unpaved shoulders (Figure E.14 and Figure E.15). Both shoulder types were evalu- ated together as part of the quality assurance process. The indi- cation of paved or unpaved shoulder was first verified for accuracy. The type of shoulder for both the unpaved and paved Figure E.3. Pennsylvania 2013 grade random and control site chart (location).
96 0 2 4 6 8 10 12 14 16 18 20 Pennsylvania 2013 Grade Comparison Ctre Fugro Figure E.4. Pennsylvania 2013 random site grade comparison. Figure E.5. Pennsylvania 2013 cross slope random and control site charts. Figure E.6. Pennsylvania 2013 intersection random and control site charts. Figure E.7. Pennsylvania 2013 speed signs random and control site charts.
97 Figure E.8. Pennsylvania 2013 signs random and control site charts. Figure E.10. Pennsylvania 2013 rumble strips random and control site charts. Figure E.9. Pennsylvania 2013 highway lighting random site chart. shoulders was then checked. The paved shoulders required an additional step to evaluate the width. No widths were collected as part of the unpaved shoulder data. For the control sites, mobile-collected curve point of cur- vature (PC), point of tangency (PT), radius, and length were compared to the corresponding department of transportationâ provided alignment attributes. Random curve evaluation required a separate process, because these data could not be accurately found in the field data set. The process used GPS traces from the mobile data collection to measure the radius of curves. By selecting points within a curve, the length of curve and the chord of the curve were calculated. Using these values, the curve radius could accurately be calculated and compared to the mobile vendor radius values. The GPS trace points were provided at 21 ft intervals along the roadways collected. The research team plotted these points in Esri ArcGIS, along with select points within random curves across the entire network (Figure E.16). An equal proportion of curves in the ranges of less than 1,500 ft, 1,500 ft to 6,000 ft, and greater than 6,000 ft were selected. Because the radius of the curve was not known when selecting the curve, multiple iterations were completed until the desired number of curves was found in each range. Once the points were selected, a polyline was created by joining the GPS traces of the curve. From this polyline, the length of curve and chord of the curve were determined. The radius of the curve was derived from these values and spa- tially joined to the respective mobile data curve. Figures E.17 to E.19 show the results after the comparison of the radii were completed. Any curves that were identified as problems were further evaluated for discrepancies. This concluded the quality assurance process. Any con- firmed errors were reported in detail and provided to the contractor to address the issues with the data. Any identi- fied problem was remediated through correction in post- processing of the data or, in rare occasions, re-collection of the data. The majority of the corrections took place during the pilot phase of the data collection.
98 Figure E.12. Pennsylvania 2013 guardrail random and control site charts. Figure E.14. Pennsylvania 2013 paved shoulder random and control site charts. Figure E.13. Pennsylvania 2013 lanes random and control site charts. Figure E.11. Pennsylvania 2013 medians random site chart.
99 Figure E.17. Pennsylvania 2013 radius range of less than 1,500 ft. Figure E.15. Pennsylvania 2013 unpaved shoulder random and control site charts. Figure E.18. Pennsylvania 2013 radius range of 1,500 ft to 6,000 ft. Figure E.16. Selection of GPS traces for curve.
100 Figure E.19. Pennsylvania 2013 radius range of greater than 6,000 ft.
