Potato crop stress identification in aerial images using deep learning?based object detection

Recent research on the application of remote sensing and deep learning-based analysis in precision agriculture demonstrated a potential for improved crop management reduced environmental impacts agricultural production. Despite promising results, practical relevance these technologies field deployment requires novel algorithms that are customized images robust to implementation natural imagery. The paper presents an approach analyzing aerial potato (Solanum tuberosum L.) using neural networks. main objective is demonstrate automated spatial recognition healthy vs. stressed at plant level. Specifically, we examine premature senescence resulting drought stress Russet Burbank plants. We propose learning (DL) model detecting stress, named Retina-UNet-Ag. proposed architecture variant Retina-UNet includes connections from low-level semantic representation maps feature pyramid network. also introduces dataset acquired with Parrot Sequoia camera. manually annotated bounding boxes regions. Experimental validation ability distinguishing plants images, achieving average dice score coefficient (DSC) 0.74. A comparison related state-of-the-art DL models object detection revealed presented effective this task. method conducive toward assessment collected under conditions.