Radiometric correction of terrestrial LiDAR point cloud data for individual maize plant detection

Abstract

Detailed geoinformation on in-field variations of plant properties (e.g., density, height) is required in precision agriculture and serves as a valuable input for plant growth models and crop management strategies. This letter presents a novel workflow for object-based point cloud analysis for individual maize plant mapping, using radiometric and geometric features of terrestrial laser scanning. The performed radiometric correction achieves a reduction of amplitude variation of homogeneous areas to 1/3 of the original variation and offers a distinct separability of the target class maize plant from soil. The developed procedure, including 3-D point cloud filtering and segmentation, is able to reliably detect single plants with a completeness {>}{\rm 80}\% and correctness {>}{\rm 90}\%. Experiments on reduced point densities show stability of detection rates above 100 points per 0.01 {\rm m}2. The results indicate that the developed workflow will lead to even higher detection accuracy with LiDAR point clouds captured by mobile platforms, with less occlusion effects and more homogeneous point density. © 2013 IEEE.