Threshold-Based Automated Pest Detection System for Sustainable Agriculture

TL;DR

This paper introduces a threshold-based IoT and computer vision system for automated pea weevil detection, aiming to improve sustainable agriculture by managing pest populations effectively and scalably.

Contribution

It presents a novel, non-machine learning approach using thresholding and contour detection for pest identification, integrated into a scalable IoT system for sustainable farming.

Findings

Effective weevil detection demonstrated in tests

System scalable for resource-constrained environments

Open-source software available for research community

Abstract

This paper presents a threshold-based automated pea weevil detection system, developed as part of the Microsoft FarmVibes project. Based on Internet-of-Things (IoT) and computer vision, the system is designed to monitor and manage pea weevil populations in agricultural settings, with the goal of enhancing crop production and promoting sustainable farming practices. Unlike the machine learning-based approaches, our detection approach relies on binary grayscale thresholding and contour detection techniques determined by the pea weevil sizes. We detail the design of the product, the system architecture, the integration of hardware and software components, and the overall technology strategy. Our test results demonstrate significant effectiveness in weevil management and offer promising scalability for deployment in resource-constrained environments. In addition, the software has been open-sourced for the global research community.