The use of LiDAR to identify forest transportation networks

by Storm J. Beck

Institution: Oregon State University
Department: Sustainable Forest Management
Degree: MS
Year: 2014
Keywords: LiDAR; Forest products  – Transportation  – Remote sensing
Record ID: 2030370
Full text PDF: http://hdl.handle.net/1957/46926


The production of high value non-conventional products, such as long utility poles; or the production of low value bulky products, such as chips or grindings; provide opportunities for forest owners to increase value from their forests. The transport of these products requires the use of specialized trucks and trailers. However, the lack of engineering records of forest roads provides a challenging environment in the assessment of transportation of non-conventional products. The primary challenge to transporting non-conventional products is determining if the specialized vehicle can navigate the horizontal and vertical geometry, as well as turning around near the landing. LiDAR provides data that could aid in the evaluation of specialized vehicles at the transportation network scale. In this thesis, a review of previous research using aerial and terrestrial LiDAR to identify the forest transportation network is made. From this review it was evident that few studies have tried to automatically extract forest road location. Hence, a process to identify and extract forest roads from a LiDAR data was developed and implemented. The two main principles that were used to identify forest roads were (1) intensity values change with material properties and (2) ground point densities differ on forest roads compared with the forest floor. These two principles are used in conjunction with buffering, removing, and connection routines. The removing and connection routines work to remove short isolated road segments and to connect segmented road segments. The road extraction process identified 67 percent of the roads that were field sampled. If gravel and native surface roads were separated from the analysis, the process identified 84 percent of the gravel and 10 percent of the dirt forest road segments by length. When assessing results of the road extraction process across the entire area stratified by canopy cover, the results were 80 percent true positives, 34 percent false positives, 20 percent false negative, 38 percent true negatives. Finally the road geometry of the aerial LiDAR data were compared to terrestrial LiDAR data. This comparison focused on the following attributes, road width, cross-slope, left cut/fill slope, right cut/fill slope. The average absolute difference in the road width between the two methods was 1.1m, the cut/fill slope differences was less than four percent, and the difference in road cross slope was two percent. These results are comparable to other published results. Future research and additions to this road identification and extraction process include adding an image analysis process to help identify roaded areas and eliminate large areas of non-roaded area as identified in the first process. After the road identification process a thinning algorithm could be used to identify the road centerline providing vehicle paths throughout the transportation network. These paths and a 3-D model of the forest road could be used in a vehicle conflict analysis. Finally, a…