Abstract
Extraction of leaf area index (LAI) is an important prerequisite in numerous studies related to plant ecology, physiology and breeding. LAI is indicative for the performance of a plant canopy and of its potential for growth and yield. In this study, a novel method to estimate LAI based on RGB images taken by an unmanned aerial system (UAS) is introduced. Soybean was taken as the model crop of investigation. The method integrates viewing geometry information in an approach related to gap fraction theory. A 3-D simulation of virtual canopies helped developing and verifying the underlying model. In addition, the method includes techniques to extract plot based data from individual oblique images using image projection, as well as image segmentation applying an active learning approach. Data from a soybean field experiment were used to validate the method. The thereby measured LAI 14 prediction accuracy was comparable with the one of a gap fraction-based handheld device (R2 of 0.92, RMSE of 0.42 m2 m2) and correlated well with destructive LAI measurements (R2 of 0.89, RMSE of 0.41 m2 m2). These results indicate that, if respecting the range (LAI ≤3) the method was tested for, extracting LAI from UAS derived RGB images using viewing geometry information represents a valid alternative to destructive and optical handheld device LAI measurements in soybean. Thereby, we open the door for automated, high-throughput assessment of LAI in plant and crop science.
Leaf area index estimation using top-of-canopy airborne RGB images
