Optimization of soil background removal to improve the prediction of wheat traits with UAV imagery

Grain yield and quality assessment based on remote sensing throughout the growth cycle may contribute to achieve efficient sustainable wheat production. A trial was conducted test suitability of a multispectral sensor onboard an unmanned aerial vehicle (UAV) estimate yield, protein content, nitrogen (N) output in winter (Triticum aestivum L.). Furthermore, method improve predictions traits proposed by sequentially removing soil pixels from distribution vegetation index (VI) values. The experimental area, located at Aranjuez (Madrid, Spain), divided into four sectors with 133 plots total. Different N doses were applied create high intravariability. In each plot, measured harvest. An flown stages (GS) – GS32, GS39, GS65 ensure correct monitoring crop. Four VIs, Normalized Difference Vegetation Index (NDVI), Modified Soil (MSAVI), Red Edge (NDRE) Blue-Red (BRI1) selected spectral information their for estimating traits. sequential cutting (threshold value optimization, TVO) implemented remove background pixels, threshold values computed VIs distributions. Then, predictive performance segmentation evaluated. Our results indicated that NDVI, MSAVI, NDRE able predict using sensors UAV. We optimal thresholds ranging 0.1 0.3 depending VI trait. TVO showed improvement estimation stem elongation stage (GS32). However, achieved limited content anthesis (GS65). Overall, our suggest (a) reflectance is essential element UAV imagery introduces uncertainty grain (b) mitigate effect.