Applications of unmanned aerial vehicle (UAV) spectral systems in precision agriculture require raw image data to be converted to reflectance to produce time-consistent, atmosphere-independent images. Complex light environments, such as those caused by varying weather conditions, affect the accuracy of reflectance conversion. An experiment was conducted here to compare the accuracy of several target radiance correction methods, namely pre-calibration reference panel (pre-CRP), downwelling light sensor (DLS), and a novel method, real-time reflectance calibration reference panel (real-time CRP), in monitoring crop reflectance under variable weather conditions. Real-time CRP used simultaneous acquisition of target and CRP images and immediate correction of each image. These methods were validated with manually collected maize indictors. The results showed that real-time CRP had more robust stability and accuracy than DLS and pre-CRP under various conditions. Validation with maize data showed that the correlation between aboveground biomass and vegetation indices had the least variation under different light conditions (correlation all around 0.74), whereas leaf area index (correlation from 0.89 in sunny conditions to 0.82 in cloudy days) and canopy chlorophyll content (correlation from 0.74 in sunny conditions to 0.67 in cloudy days) had higher variation. The values of vegetation indices TVI and EVI varied little, and the model slopes of NDVI, OSAVI, MSR, RVI, NDRE, and CI with manually measured maize indicators were essentially constant under different weather conditions. These results serve as a reference for the application of UAV remote sensing technology in precision agriculture and accurate acquisition of crop phenotype data.

Lodging is a common problem in maize production that seriously impacts yield, quality, and the capacity for mechanical harvesting. Evaluation of site-specific lodging risks requires establishment of a method for multi-year monitoring. In this study, spectral images collected by the Sentinel-2 satellite were processed to obtain three types of data: gray-level co-occurrence matrix texture (GLCM), vegetation indices (VIs), and spectral reflectance (SR). Lodging classification models were then established with Random Forest (RF) using each of the three data types separately (the GLCM, VI, and SR models) and in combination (SR+VI model, SR+GLCM model, VI+GLCM mod-el, and SR+VI+GLCM model). By gradually removing features with low importance scores from the SR+VI+GLCM model and analyzing the changes in the overall accuracy (OA), the optimal set of predictive variables was identified and used to construct the optimal model. A model built us-ing data from a single timepoint in 2021 was tested on data collected at a similar timepoint in 2019 and vice versa to assess interannual model generalizability. The results of this study demon-strate that for monitoring maize lodging, models constructed with a single feature type, the GLCM model had significantly lower accuracy compared to the VI and SR models. During certain growth stages, the model constructed with combined features had significantly higher accuracy in monitoring maize lodging compared to models constructed with a single feature. During the pro-cess of selecting the optimal predictive variables, it was found that the accuracy of the model did not increase as the number of predictive variables increased. The results show that the positive and negative validation models had an accuracy of 96.55% and 95.18%, with kappa values of 0.93 and 0.83, respectively. This indicates that the model has strong generality for the same repro-ductive stage between years. This study provides a detailed method for large-scale maize lodging monitoring, allowing for identification of optimal planting practices to reduce the probability of lodging and ultimately improving regional maize yield and quality.