Assessing the potential for bicycle-assisted cross-slope estimation



Nault, Andrew

Journal Title

Journal ISSN

Volume Title



The objective of this thesis is to assess the suitability of using both raw and processed USGS LiDAR data to determine walkway cross-slope using. Areas with greater than 2% cross-slope exceed the accessibility design guidelines for sidewalk crossslope published as a part of the Americans with Disabilities Act (ADA). A GIS (Geographic Information System) algorithm to produce cross-slopes from LiDAR (Light Imaging, Detection, And Ranging) derived DEMs (Digital Elevation Maps) was utilized for this work, and its accuracy was compared to ground-truth cross-slope from individual LiDAR pules as well as readings from a hand-held digital inclinometer. A secondary objective is to utilize image recognition analysis to identify different the types of walkway surface material used along the surveyed walkway. A tertiary objective is to use a bicycle mounted GoPro camera system capture mono and stereo imagery in conjunction with an iPad or Android-GPS-cellphone used to capture position. In a MATLAB processing environment, GoPro imagery was corrected for lens distortion in both the stereo and mono versions. Imagery location was determined by assigning individual GoPro images (or stereo-pairs) to GPS points in ArcGIS using time as the primary table key. Analyst-derived edge detection was used to find width and then find cross-slope (for the stereo-pairs only) from the 3D stereo-pair point clouds. In addition to building a georeferernced library of images for assessment of walkway condition, There were multiple technical results from this thesis research. It was found that the accuracy of the GPS receiver telemetry was dependent on the presence or absence of cellular-network-geopositional accuracy enhancement subroutines as well as vegetative canopy cover. In terms of cross-slope results it was found that in the context of Fairfax County’s Accotink multi-use trail, there was a greater overall incidence of > 2% crossslope than Poquoson’s Tidewater, marsh-based sidewalk network. DEM accuracy was assessed with two methodologies, with digital inclinometer and by picking out individual LiDAR transect points from a raw point-cloud and using the slope intercept formula. It was found that LiDAR based DEMS underestimated slope as compared to the two ground-truthing methods. The surface classification methods were effective in determining gravel and concrete but less effective in determining the presence of other surface types. This thesis consisted of three phases and occurred in two geographic locations. Phase one and two occurred in Fairfax County’s Accotink Creek multi-use trail. Phase two utilized a cell-network accuracy assisted GPS while phase two did not. The third phase occurred on walkways in the city of Poquoson, Virginia.



Cross-slope, LiDAR, GIS, Bicycle, Assessment, Stereoscopic