Remote sensing of trees using full waveform LiDAR data and optical imagery: A case study of tree counting in Madera, California

Abstract

The study explores the feasibility of combining full-waveform LiDAR data with optical imagery to accurately count trees by simultaneously analyzing the structural characteristics of both data types. A new approach was developed to demonstrate the potential of full-waveform LiDAR data in tree counting. The method involved generating a binary vegetation map from an aerial color image and segmenting it. Conventional features for the LiDAR point cloud, and 11 structural features for the LiDAR waveforms were generated. Using generated features, a multi-step filter-out process was applied to remove extra points and identify the number of trees in each segment. The method achieved a tree-counting accuracy of about 94%, indicating reliable results. Further analysis revealed a 17% commission error and an 11% omission error resulted from overestimating and underestimating trees, respectively. To compare the strength of data integration, tree counting was performed independently on waveform LiDAR and optical image independently. However, the accuracy of the results obtained from each data set separately was lower than that obtained from integrating both data sets. Errors were more prevalent in large segments with over five trees or dense tree cover of different shapes and sizes. The integration of full-waveform LiDAR data enhanced accuracy in tree counting, overcoming the limitations of individual data sources. The study highlights the significance of integrating these data types for improved remote sensing applications. © 2025 Elsevier B.V.