Pengcheng Hu1, 2, Tao Duan1, 2, Scott Chapman2, 3, Yan Guo1 and Bangyou Zheng2*
1 College of Resources and Environmental Sciences, China Agricultural University, Beijing 100193, China
2 CSIRO Agriculture and Food, Queensland Biosciences Precinct 306 Carmody Road, St Lucia, 4067, QLD, Australia
3 School of Food and Agricultural Sciences, The University of Queensland, via Warrego Highway, Gatton 4343, QLD, Australia
Corresponding Author: Bangyou Zheng, Tel: +61 (0)7 3214 2620, Fax: +61 (0)7 3214 2920, email: Bangyou.Zheng@csiro.au
Canopy height is a simple trait to represent status of plant growth and development and potentially biomass production, as well as having an influence on lodging susceptibility. Measurement is labour-intensive and time consuming, especially in the large breeding trials, which involve thousands of plots. In this study, we developed a methodology to estimate canopy height using a high resolution camera which was mounted on an Unmanned Aerial Vehicle (UAV). Aerial images were captured during a wheat growth season with contrasting canopy height at four time points. Digital surface models (DSMs) were generated using structure from motion (SfM) algorithm after 3D reconstruction of the whole field (e.g. phenocopter.csiro.au). Three methods were used to estimate soil surface and canopy heights at plot level after segmentation of DSM into individual plots including 1) background extraction after segmentation mosaic into vegetation; 2) interpolation of soil surface using soil surface around field edges; 3) measured soil DSM by a flight immediately after planting. Estimates of canopy heights were compared to manual measurements close to the flight times. The results indicated Method 2 had the best performance for wheat canopy (R2 = 0.88; RMSE = 4.5 cm). The proposed method can been integrated into the high throughput phenotyping platform to apply UAV technologies in the breeding program.