基于人工智能的农田景观中害虫发生预测与展望
Prediction and prospects of pest occurrence in agricultural landscapes based on artificial intelligence
杨璐嘉** 门兴元***
点击：128次 下载：8次
DOI：10.7679/j.issn.2095-1353.2025.054
作者单位：山东省农业科学院植物保护研究所，山东省农业有害生物绿色防控重点实验室，济南 250100
中文关键词： 农田景观；害虫预测；人工智能；时空建模；深度学习
英文关键词：agricultural landscape; pest prediction; artificial intelligence; space-time modeling; deep learning
中文摘要：
农田景观中害虫发生、扩散和危害，导致了作物减产、农产品质量下降，并深刻影响了农业生态系统的稳定性和可持续性。作为农业生产管理的关键环节，害虫种群监测与发生预测，有助于提前预警和合理防控决策，在保障粮食安全和推动生态农业建设中发挥着至关重要的作用。传统的害虫预测方法主要利用种群调查、环境监测、景观遥感以及历史数据的统计建模，但在应对全球气候变化和农田景观变化、提高预测精准度以及实现动态监测方面仍面临诸多局限。随着人工智能（AI）技术的进步，基于机器学习和深度学习在农田景观中害虫预测系统成为研究热点。本文系统综述了农田景观中害虫预测技术的演进路径及智能预警机制的理论与应用，重点讨论了基于AI驱动在农田景观害虫预测中的应用潜力、深度学习模型在害虫智能识别中的实践应用、以及多模态数据融合技术在害虫-天敌互作监测中的潜力等方面。本文还详细分析了当前AI模型在害虫预测中可能面临的技术瓶颈与挑战，展望了通过强化跨学科合作与技术集成，AI技术对农田景观中的害虫预测的深度应用将进一步优化农业生态管理策略，提高害虫防控的精准性和时效性。
英文摘要： The occurrence, spread and damage of pests within agricultural
landscapes lead to reduced crop yields, lower quality of agricultural products,
and impacts on the stability and sustainability of agroecosystems. As a key component of agricultural management, pest population
monitoring and occurrence prediction contributes to early warning and rational
prevention and control decisions and plays an important role in ensuring food
security and promoting ecological agriculture. Traditional pest prediction
methods mainly use population surveys, environmental monitoring, landscape
remote sensing, and statistical modeling based on historical datasets, while
they still face significant limitations in responding to the change of global
climate and agricultural landscapes, improving prediction accuracy, and
achieving dynamic monitoring. With the advancements of artificial intelligence
(AI) technology, pest prediction systems based on machine learning and deep
learning in agricultural landscapes have become a hot research topic. This
study provides a systematic overviews of the evolutionary trajectory of pest
prediction technologies and the theoretical foundations and applications of
intelligent early warning mechanisms in agricultural landscapes, focusing on
the potential application of AI-driven pest prediction in agricultural
landscapes based on AI, the practical application of deep learning models for
intelligent pest recognition, and the potential of multimodal data fusion
technologies in pest-natural enemy interactions monitoring. This study also
analyzes in detail the current technical bottlenecks and challenges that AI
models may face in pest prediction, and foresees that by strengthening
interdisciplinary cooperation and technical integration, the in-deep application
of AI for pest prediction in agricultural landscapes will further optimize
agro-ecological management strategies, and improve the precision and timeliness
of pest prevention and control.