A new approach to retrieve leaf normal distribution using terrestrial laser scanners

摘要

Leaf normal distribution is an important struc-tural characteristic of the forest canopy. Although terres-trial laser scanners (TLS) have potential for estimating canopy structural parameters, distinguishing between leaves and nonphotosynthetic structures to retrieve the leaf normal has been challenging. We used here an approach to accurately retrieve the leaf normals of camphorwood (Cinnamomum camphora) using TLS point cloud data. First, nonphotosynthetic structures were filtered by using the curvature threshold of each point. Then, the point cloud data were segmented by a voxel method and clustered by a Gaussian mixture model in each voxel. Finally, the normal vector of each cluster was computed by principal compo-nent analysis to obtain the leaf normal distribution. We collected leaf inclination angles and estimated the distri-bution, which we compared with the retrieved leaf normal distribution. The correlation coefficient between measure-ments and obtained results was 0.96, indicating a good coincidence.