Introduction risk assessment for quarantine pests by environmental monitoring, object detection and Monte Carlo simulation

Abstract

Monitoring and detecting quarantine pests (Q-pests) in cross-border cargo holds is essential for preventing biological invasions. To satisfy the demands of customs for efficient inspection at entry ports, we propose an introduction risk assessment system for Q-pests in cross-border cargo holds. First, we develop a multi-sensor cargo hold monitoring system that not only traps and detects Q-pests but also tracks various cargo hold environmental parameters, such as temperature, humidity, and fumigant gas. Second, we construct a dedicated Q-pest image detection dataset, named QP-5K, and propose a unified Q-pest detection network QPNet, which significantly improves Q-pest detection performance. Finally, we develop a quantitative introduction risk assessment method by fusing multi-source information, including environmental parameters and Q-pest detection results. To address the high cost associated with obtaining extensive real monitoring data, Monte Carlo simulation is utilized to generate monitoring data and determine the distribution weights. The experimental results show that QPNet achieves state-of-the-art performance with an AP of 90.4% on the QP-5K dataset and 47.5% on the Pest24 dataset, and the Monte Carlo simulation results also verify the effectiveness of the proposed risk assessment method. In summary, the proposed system provides a more accurate and reliable approach for assessing the introduction risk of Q-pests during cross-border transportation.