Multi-source UAV remote sensing and AI for crop growth monitoring

Accurate and in-time monitoring of cropping systems is critical to precision farming in order to facilitate decision-making for agronomic management and enhancing crop yield under changing climate. In this study, multi-source unmanned aerial vehicle (UAV) remote sensing observations were conducted at several key growing stages of crops at a standard wheat-maize cropping system field trials in the North China Plain from 2018 to 2020. Crop leaf area index, above-ground biomass, chlorophyll content, grain yield, and plant density were estimated using multi-source UAV remote sensing observations (including RGB, multi/hyperspectral, LiDAR, and thermal sensors) processed by machine/deep learning approaches.

In this study, we will give a comprehensive research introduction focusing on how to improve the estimation accuracy of the above crop growth variables via UAV remote sensing and machine/deep learning approaches, including three aspects:

(1) Data source and fusion, including the integration of multi-source UAV information for comprehensive maize growth monitoring, comparison of UAV-based point clouds with different densities for crop biomass estimation, and crop chlorophyll content estimation using multi-scale hyperspectral information.

(2) Optimization of UAV observation management: we will answer when is the most relevant phenological stage for maize yield estimation via high-frequent UAV observations; investigate extrapolation artefacts, validate the suitability and discuss the uncertainty of the UAV-based strategies for 'model calibration at a small site while applying these models at a large extent' for crop monitoring.

(3) Modeling improvement will give two cases to introduce improving crop biomass estimation accuracyand realize the plant density counting during the vigorous growing period employing deep learning.