农业害虫自动识别与监测技术
摘要
隨着计算机和互联网技术的发展,信息技术已被广泛地应用于植物保护领域,推动农业害虫的监测走向信息化、智能化和精准化。我们综述了农业害虫自动识别与监测技术的最新研究进展,分析了各种技术的特点与优势。这些技术均需要特定的设备获取农业害虫及其生境的信息,提取昆虫信息特征,并利用这些特征进行昆虫种类的识别与计数,达到害虫监测的目的。图像识别技术适合于自动识别与监测栖息于作物表面的害虫,昆虫雷达(厘达或激光雷达)技术特别适合于自动识别与监测高空中飞行的害虫,而声音识别技术在自动识别与监测隐蔽害虫方面具有优势。最近发展起来的基于深度学习的害虫识别方法避免了传统的手工设计特征方法,提高了害虫识别的鲁棒性,展示了一旦建立完整的昆虫信息库就可以实现害虫自动识别与监测的可能;这给昆虫学家提出了一个艰巨的任务,即鉴定和正确标识机器学习所需的大量的昆虫信息。
关键词
农业害虫; 监测技术; 图像识别; 声音识别; 昆虫雷达; 深度学习
中图分类号:
S 431.9
文献标识码: A
DOI: 10.16688/j.zwbh.2018305
Automatic identification and monitoring technologies of
agricultural pest insects
FENG Hongqiang YAO Qing2
(1. Henan Key Laboratory of Crop Pest Control, MOA’s Regional Key Lab of Crop IPM in Southern Part of
Northern China, Institute of Plant Protection, Henan Academy of Agricultural Sciences, Zhengzhou 450002, China;
2. College of Information, Zhejiang Sci-Tech University, Hangzhou 310018, China)
Abstract
With the development of computer and internet technology, information technology has been applied widely in plant protection. It promotes the monitoring of agricultural pest insects towards informatization, intelligence and precision. We reviewed recent research progress of automatic identification and monitoring of pests and analyzed the features and advantages of each technology. All technologies need to acquire information of insects and their environments, extract insect features and automatically identify insect species and count pests for monitoring pests. The image identification technology is suitable for identification and monitoring of pests resting on surface of crops or trapped pests. Entomological radar (or lidar) has advantages in identification and monitoring of insects freely flying at high altitudes. Acoustic technology has advantages in identification and monitoring hidden pests. Recently, deep learning approach has been reported to achieve state-of-the-art performance on object recognition tasks. It improves the robustness of pest identification without prior knowledge and hand effort in feature design. It means its possibility of automatic monitoring of pest insects once we set up a database of insect information. This gives entomologists a hard task for identifying and accurately labelling huge number of insect information for machine learning.
Key words
agricultural pest; monitoring technology; image identification; acoustic recognition; entomological radar; deep learning