PestDSS: An Integrated Decision Support System for Sustainable Pest Management in Agriculture

The direct and indirect damage to crops caused by pests is a major factor affecting crop yields. Providing farmers with professional and cost-effective pest management decisions in a timely and accurate manner is a challenge in precision agriculture. Currently, researchers propose agricultural decision support systems using database and data analysis techniques to provide farmers with expert support and economic thresholds for pest management. However, these efforts overlook the challenge of identifying multiple pest species for agricultural workers and human error in the manual monitoring of pest densities. We propose PestDSS, an object detection-based decision support system to address the aforementioned challenges, which integrates agricultural decision support systems and state-of-the-art object detection models to semi-automatically make pest management decisions for farmers. Specifically, PestDSS includes a cloud-based farm information management module that allows users to manage their own farm data, a knowledge base of agriculture module, and a decision-making tool module for pest management. The decision-making tool combines outputs of an object detection model with optimisable thresholds developed through expert knowledge to provide pest management decisions. The proposed pest detection model outperforms current state-of-the-art object detection models on three pest detection datasets. We apply PestDSS to a case study of wheat pest management to demonstrate the usability of the system and illustrate the potential for its use.