@inproceedings{625,
  author = {Roh Detor S.},
  title = {Analyzing Domain Knowledge for Big Data Analysis: A Case Study with Urban Tree Type Classification.},
  abstract = {The goals of this research were to create a labeled dataset of tree shadows and to test the feasibility of shadow-based tree type identification using aerial imagery. Urban tree big data that provides information about individual trees can help city planners optimize positive benefits of urban trees (e.g., increasing wellbeing of city residents) while managing potential negative impacts (e.g., risk to power lines). The continual rise of tree type specific threats, such as emerald ash borer, due to climate change has made this problem more pressing in recent years. However, urban tree big data are time consuming to create. This paper evaluates the potential of a new tree type identification method that utilizes shadows in aerial imagery to survey larger regions of land in a shorter amount of time. This work is challenging because there are structural variations across a given tree type and few verified tree type identification datasets exist. Related work has not explored how tree structure characteristics translate into a profile view of a tree’s shadow or quantified the feasibility of shadow-only based tree type identification. We created a consistent and accurate dataset of 4,613 tree shadows using ground truthing procedures and novel methods for ensuring consistent collection of spatial shadow data that take binary and spatial agreement between raters into account. Our results show that identifying trees from shadows in aerial imagery is feasible and merits further exploration in the future.},
  year = {2019},
  journal = {Big Data Analytics. BDA 2019. Lecture Notes in Computer Science},
  volume = {11932},
  month = {01},
  url = {https://par.nsf.gov/biblio/10184558},
  doi = {10.1007/978-3-030-37188-3_11},
}