Delay Efficient Federated Learning based Plant Disease Detection and Monitoring (DFLPDDM) in Agricultural Fields: A UAV-IoT Environment

Abstract

Plant disease detection, monitoring and smart spraying using Unmanned Aerial Vehicle-Internet of Things (UAV-IoT) environment, has become extremely important in precision agriculture, this has to be performed in delay efficient manner so that prompt action can be taken to save as many plants as possible. Energy efficiency is an added advantage that incorporates sustainability in thesolution. However, to the best of authors knowledge, articles in relevant literature have neglected the concept of security which is indispensable if different regions of a big agricultural land belong to different owners and they want to keep their land status information confidential. In this article, we propose a delay efficient federated learning-based plant disease detection and monitoring scheme (DFLPDDM) that utilizes the concept of transmitting gradients among untrusted UAV’s and transmitting data among trusted UAVs to ensure security and confidentiality. Also, mechanisms are proposed to incorporate delay and energy efficiency to improve overall performance effectiveness of the system. Simulation results shows that DFLPDDM produce much better performance compared to many other state-of-the-art agricultural field monitoring algorithms.

Plant phenotyping relevance

植物病害状態の検出・監視を対象に、UAV-IoTと連合学習を組み合わせた技術を提案しており、病害フェノタイプの取得・推定手法が中心である。

Code and data availability

The paper describes a purely simulated UAV-IoT federated learning scheme (Python framework, synthetic parameters in Table 1) with no public phenotype/trait datasets, plant images, sensor data, author code repository, trained models, or supplements. No data or code availability statement appears anywhere in the supplied

No evidence-backed public reproduction asset is currently recorded.