UAVEC-FLchain: Distributed multi-regional jujube orchard joint yield estimation for secure agricultural-IoT applications

In recent years, Internet of Things (IoT) is closely integrated with smart agriculture, providing better technical guarantee for the growth management and yield estimation of agricultural products. In this paper, targeted at the distributed jujube orchards, we develop a federated learning-based framework for secure multi-regional yield estimation on IoT edge devices. In specific, we propose the UAVEC-FLchain framework for secure and efficient jujube yield estimation, consisting of the UAVEC module, data joint processing module, and data tracking module. Different UAV devices are used at edge terminals to collect distributed images of different jujube orchards, and the model is trained securely with federation learning and blockchain. Furthermore, we propose a method to avoid malicious node attacks by calculating the contribution of computing nodes. Experimental results indicate that as the number of participating nodes increases, the average detection accuracy improves. The results of training with 60 nodes show an 8.06 % improvement over training with 6 nodes. When facing malicious node attacks, the relative error of the proposed framework is 10.24 % lower than that with traditional federated learning.

Journal

INTERNET OF THINGS

Year

2024

Volume

25

ISSN

2543-1536

eISSN

2542-6605

DOI

10.1016/j.iot.2024.101143

WoS UID

WOS:001215096100001