3D Dynamic Morphological Structure of a Crop Captured from a UAV.

The development of reliable methods for estimating 3D plant morphological structure during a crop growing season is a key issue for the quantitative evaluation of crop condition and response. Characterizing structural measurements of plant properties is of interest in stress detection and crop yield estimation, since crop responses are strongly correlated to their 3D morphology. With improvements in miniaturized cameras, Unmanned Aerial Vehicle (UAV) technologies and the support of advanced computer vision and photogrammetric algorithms, routine monitoring of plant features is now feasible, offering spatially explicit retrievals on an as needed basis. Using Structure from Motion (SfM) techniques on collected high-resolution RGB imagery allows for the reconstruction of point clouds, and hence Digital Surface Maps (DSM), from which the dynamics of crop structure and development over time can be determined. Here we evaluate a UAV-based SfM technique for dynamic 3D crop monitoring of tomato plants and assess a methodology for estimating plant structure parameters (volume, leaf area and crop height) from point clouds generated using SfM. High-resolution optical imagery was collected over 10 flights during a 3 month growing season using a quadcopter drone. A geographical information system analysis and object-based classification were used for the calculation of crop parameters. UAV retrievals were compared against terrestrial lidar scans for the purpose of evaluating the ability of SfM point clouds to correctly capture the morphology of the tomato plants. Results show that the approach is able to accurately reproduce the 3D information of the tomato plants throughout the duration of the growing season. Overall, it was found that 3D plant reconstruction at different growth stages provides a valuable health indicator, providing useful information with which to inform precision agricultural management at the local scale.