NEON NIST data science evaluation challenge: methods and results of team Conor

• DOI: 10.7287/peerj.preprints.26977v1
• Publication Date: 2018-06-04T13:23:38Z
• Authors: Conor A McMahon
• Keywords: NIST; Identification; Data set; Tree (set theory)

The NIST DSE Plant Identification challenge is a new periodic competition focused on improving and generalizing remote sensing processing methods for forest landscapes. To compete in the competition, I created pipeline to perform three tasks. First, NDVI- height-thresholded watershed segmentation was performed identify individual tree crowns using LIDAR height measurements. Second, data segmented aligned with ground measurements by choosing set of pairings which minimized error position crown area as predicted stem height. Third, species classification reducing dataset's dimensionality through PCA then constructing maximum likelihood classifiers estimate likelihoods each tree. Of algorithms, routine exhibited strongest relative performance, algorithm performing least well.