Search

Background: The tiller number per unit area is one of the main agronomic components in determining yield. A real-time assessment of this trait could contribute to monitoring the growth of wheat populations or as a primary phenotyping indicator for the screening of cultivars for crop breeding. However, determining tiller number has been conventionally dependent on tedious and labor-intensive manual counting. In this study, an automatic tiller-counting algorithm was developed to estimate the tiller density under field conditions based on terrestrial laser scanning (TLS) data. The novel algorithm, which is named ALHC, involves two steps: (1) the use of an adaptive layering (AL) algorithm for cluster segmentation and (2) the use of a hierarchical clustering (HC) algorithm for tiller detection among the clusters. Three field trials during the 2016–2018 wheat seasons were conducted to validate the algorithm with twenty different wheat cultivars, three nitrogen levels, and two planting densities at two ecological sites (Rugao & Xuzhou) in Jiangsu Province, China. Result: The results demonstrated that the algorithm was promising across different cultivars, years, growth stages, planting densities, and ecological sites. The tests from Rugao and Xuzhou in 2016–2017 and Rugao in 2017–2018 showed that the algorithm estimated the tiller number of the wheat with regression coefficient (R2) values of 0.61, 0.56 and 0.65, respectively. In short, tiller counting with the ALHC generally underestimated the tiller number and performed better for the data with lower plant densities, compact plant types and the jointing stage, which were associated with overlap and noise between plants and inside the dense canopy. Conclusions: Differing from the previous methods, the ALHC proposed in this paper made full use of 3D crop information and developed an automatic tiller counting method that is suitable for the field environment. [ABSTRACT FROM AUTHOR]