Exploring tree species signature using waveform LiDAR data

Abstract:

Successful classification of tree species with waveform LiDAR data would be of considerable value to estimate the biomass stocks and changes in forests. Current approaches emphasize converting the full waveform data into discrete points to get larger amount of parameters and identify tree species using several discrete-points variables. However, ignores intensity values and waveform shapes which convey important structural characteristics.

The overall goal of this study was to employ the intensity and waveform shape of individual tree as the waveform signature to detect tree species. The data was acquired by the National Ecological Observatory Network (NEON) within 250*250 m study area located in San Joaquin Experimental Range.

Specific objectives were to: (1) segment individual trees using the smoothed canopy height model (CHM) derived from discrete LiDAR points; (2) link waveform LiDAR with above individual tree boundaries to derive sample signatures of three tree species and use these signatures to discriminate tree species in a large area; and (3) compare tree species detection results from discrete LiDAR data and waveform LiDAR data.

An overall accuracy of the segmented individual tree of more than 80% was obtained. The preliminary results show that compared with the discrete LiDAR data, the waveform LiDAR signature has a higher potential for accurate tree species classification.