Estimation of leaf area of sweet cherry trees trained as spindle using ground based 2D mobile LiDAR system

Application of tree growth models rely on the accuracy of leaf area (LA) analysis.
This study was aiming the development of a method for fast and reliable LA and volume profile estimation in sweet cherry trees grown in a commercially managed orchard.
A 2D mobile light detection and range (LiDAR) laser scanner system (MLS) was applied to obtain high density continuous tree scanning data as a 3D point cloud.
Five-year-old sweet cherry trees were scanned from both sides with the MLS equipped with a global navigation satellite system and an inertial measurement unit.
The MLS was mounted on a tractor at half of the average tree height and driven along the tree rows with 1.5 m distance from the tree stems.
MLS data acquisition was carried out before and after defoliation of the trees (n=4). The harvested leaves were manually scanned in the laboratory for calibrating the 3D point clouds.
Data processing captured merging of the generated tree point clouds by implementing an iterative closest point algorithm.
The backscattered points produced from ground were detected and segmented by implementing cloth simulation filter.
Subsequently, a volume estimation approach was proposed that segmented the tree point cloud in horizontal slices to determine the boundaries of each slice.
The LA estimated with the LiDAR using either number of laser hits per tree or estimated volume resulted in >99% accuracy for both cases, whereas the coefficients of determination were 0.92 and 0.90, respectively.
Summarizing, the leaf area analysis from 3D point cloud enabled the non-destructive evaluation of LA in cherry orchard.

IV International Symposium on Horticulture in Europe – SHE2021

K.K. Saha, N. Tsoulias, M. Zude-Sasse

laser proximal sensing, point cloud, canopy profiling

https://doi.org/10.17660/ActaHortic.2021.1327.57