S18 - Session P1 - Multi-scale image and feature field pea dataset fusion for remote sensing applications in phenomics

Authors: Sindhuja Sankaran *, Afef Marzougui, Rebecca McGee, Stephen Van Vleet

In this study, we compared the features extracted from satellite and UAS imagery to predict the seed yield from two plot-scale field pea yield trials (advanced breeding and variety testing) using the random forest model. The multi-modal and multi-scale data fusion approaches were evaluated in terms of improvement in model accuracies across trials. The multi-scale data fusion approaches also involved image fusion (pan-sharpening of satellite imagery with UAS imagery) and feature fusion (combining extracted spectral features). The main findings can be summarized as follows: texture features did not improve the model performance; the performance of the model using spectral features from satellite imagery (original resolution) can provide similar results as UAS imagery, depending on the trial and the crop growth stage; the model performance improved after multi-scale feature fusion; and the features extracted from the pan-sharpened imagery showed better model performance than those with original satellite imagery. The results from this study highlight the potential of high-resolution satellite imagery and data fusion approaches for phenotyping applications in crop research.