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

Abstract

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