Fusarium head blight (FHB) is a global wheat and barley disease that causes severe and devastating yield and quality losses. Deoxynivalenol produced by the FHB-associated pathogen is injurious to human and animal health. Hard red winter and spring wheat is a primary cereal crop in the Northern Great Plains (ND, SD, MN, NE), grown on nearly 10 million acres. FHB causes significant losses estimated at $20 million just in the South Dakota during a single year.

Therefore, developing high-yielding wheat varieties with resistance to FHB will increase crop productivity to wheat producers in the region. However, breeding for FHB resistant cultivars requires evaluating many genotypes in the field and research plots each year. Therefore, the main objective of this project was, to implement an intelligent deep scanning system that captures images from an optimum close-up to improve the FHB detection rates; Second, to develop a model using artificial intelligence and deep scanning data to enhance the Unmanned Aerial Vehicles reliability on FHB detection; Then, evaluate, calibrate, and validate the developed model and improve the aerial-based phenotyping efficiency.

Convolutional neural networks (CNNs) are an effective method to automatically identify these diseases. Recently, different approaches have been proposed to boost CNNs’ performance. In this regard, we propose the multi-scale attention U-Net. Our main goal is to improve U-net by introducing an attention mechanism at the network bottleneck. In the proposed method, we propose to include the dual attention mechanism to adaptively emphasize the importance of both spatial and channel dimensions. To this end, using the global information of each channel, we learn the scaling coefficient to improve the object learning process.

By learning the self-attention map we impose the spatial attention map, on the feature space to adaptively emphasize the important regions. We apply the suggested dual attention in a multi-scale fashion to encourage a multi-scale learning process. We prepared 12,000 images from the SDSU spring wheat breeding FHB screening nursery near Volga, SD, to train the proposed network. 

In this project, we will evaluate the developed model. This algorithm will be calibrated and be further validated for accuracy and precision. We will use the phenotyping data and the developed algorithm to improve the quality of the aerial-based phenotyping data process. This study will introduce a new generation of remote sensing-enabled information products based on a solid foundation of well-calibrated sensors, new field calibration methodologies, and innovative algorithmic techniques. 

ACKNOWLEDGEMENT AND DISCLAIMER
Confidential- please do not share the content of the submission publicly before the 2022 National FHB Forum.This material is based upon work supported by the U.S. Department of Agriculture, under Agreement, 59-0206-2-143,59-0206-2-153, and FY22-SP-002. This is a cooperative project with the U.S. Wheat & Barley Scab Initiative. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture.